{"data":[{"slug":"vercel-ai-sdk","name":"Vercel AI SDK","tagline":"Unified TypeScript SDK for building AI applications","category":"agents","url":"https://sdk.vercel.ai","github":"https://github.com/vercel/ai","pricing":"open-source","hook":"Vercel AI SDK is the abstraction layer that makes model providers interchangeable.","whatsGood":"One SDK, every model provider. Streaming responses, tool calling, structured outputs—the AI SDK handles the details while you write against a unified interface. Swap between OpenAI, Anthropic, Google, and more without changing application code.","theCatch":"Vercel-optimized, which is great on Vercel and requires more work elsewhere. Abstraction means some provider-specific features are harder to access.","verdict":"The SDK for AI applications in the Next.js ecosystem.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","active"],"capabilities":["sdk","typescript"],"surfaces":["SDK","Self-host"],"ecosystem":["Node.js","Browser"],"licenseModel":"Apache-2.0","githubStars":27389,"starsPerWeek":154,"access":"free","archived":false,"repo":{"createdAt":"2023-05-23","pushedAt":"2026-10-18","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/vercel-ai-sdk","html":"/gems/vercel-ai-sdk","markdown":"/gems/vercel-ai-sdk.md"}},{"slug":"langchain","name":"LangChain","tagline":"The OG LLM orchestration framework","category":"agents","url":"https://langchain.com","github":"https://github.com/langchain-ai/langchain","pricing":"open-source","hook":"LangChain is the Swiss Army knife of LLM development.","whatsGood":"LangChain excels at chaining together prompts, tools, and retrievers into coherent workflows. LCEL made composition feel natural after years of callback hell. The integration ecosystem is unmatched: if a vector database or LLM provider exists, LangChain probably has a connector.","theCatch":"Abstraction layers add complexity. Simple use cases feel over-engineered, and the learning curve is steeper than rolling your own for basic RAG.","verdict":"Essential for complex LLM apps, overkill for simple ones.","addedAt":"2026-08-18","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","workflows","active","fast-stars"],"capabilities":["orchestration"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":148053,"starsPerWeek":708,"access":"free","archived":false,"repo":{"createdAt":"2022-10-17","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/langchain","html":"/gems/langchain","markdown":"/gems/langchain.md"}},{"slug":"langgraph","name":"LangGraph","tagline":"Stateful multi-actor agent workflows","category":"agents","url":"https://langchain-ai.github.io/langgraph/","github":"https://github.com/langchain-ai/langgraph","pricing":"open-source","hook":"LangGraph is the operating table for agents that need to loop, branch, and remember.","whatsGood":"The graph-based approach finally makes agentic behavior tractable. You define nodes (actions), edges (transitions), and state schemas—then the framework handles persistence, streaming, and human-in-the-loop interrupts.","theCatch":"The abstraction overhead is real. Debugging graph execution requires understanding the underlying state management, and simple agents become verbose.","verdict":"The right tool when agents need memory and branching logic.","addedAt":"2026-08-18","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","workflows","active"],"capabilities":["graphs","agents"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":43480,"starsPerWeek":261,"access":"free","archived":false,"repo":{"createdAt":"2023-08-09","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/langgraph","html":"/gems/langgraph","markdown":"/gems/langgraph.md"}},{"slug":"crewai","name":"CrewAI","tagline":"Role-playing agents that actually ship","category":"agents","url":"https://crewai.com","github":"https://github.com/crewAIInc/crewAI","pricing":"open-source","hook":"CrewAI makes multi-agent collaboration as simple as defining who does what.","whatsGood":"The mental model is delightfully simple: define agents with roles, goals, and backstories, then let them work together on tasks. The role-playing paradigm sounds gimmicky but actually works—agents with defined personas produce more consistent outputs.","theCatch":"The simplicity cuts both ways. Complex inter-agent communication patterns require workarounds, and you are locked into CrewAI execution model.","verdict":"Best DX for multi-agent systems if you embrace the paradigm.","addedAt":"2026-08-18","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","active"],"capabilities":["roles","agents"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":59853,"starsPerWeek":385,"access":"free","archived":false,"repo":{"createdAt":"2023-10-27","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/crewai","html":"/gems/crewai","markdown":"/gems/crewai.md"}},{"slug":"haystack","name":"Haystack","tagline":"Production NLP pipelines from deepset","category":"agents","url":"https://haystack.deepset.ai","github":"https://github.com/deepset-ai/haystack","pricing":"open-source","hook":"Haystack was doing RAG before RAG was cool, and it shows.","whatsGood":"The pipeline abstraction is elegant: components connect through well-defined interfaces. Haystack 2.0 rewrite made composition more intuitive without sacrificing power. The evaluation framework is a standout.","theCatch":"Haystack strength is also its constraint. The framework assumes pipeline-based architectures; more exotic agent patterns require workarounds.","verdict":"The mature choice for document-centric AI applications.","addedAt":"2026-08-18","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","rag","active"],"capabilities":["nlp","pipelines"],"surfaces":["Self-host"],"ecosystem":["Python"],"licenseModel":"Apache-2.0","githubStars":26772,"starsPerWeek":74,"access":"free","archived":false,"repo":{"createdAt":"2019-11-14","pushedAt":"2026-10-17","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/haystack","html":"/gems/haystack","markdown":"/gems/haystack.md"}},{"slug":"llamaindex","name":"LlamaIndex","tagline":"Data framework for LLM applications","category":"agents","url":"https://llamaindex.ai","github":"https://github.com/run-llama/llama_index","pricing":"open-source","hook":"LlamaIndex connects LLMs to data sources that are not just documents.","whatsGood":"The connector ecosystem is unmatched: 160+ data loaders handle everything from Notion to Snowflake. The indexing abstractions (vector, keyword, knowledge graph) let you pick the right retrieval strategy for your data shape.","theCatch":"The abstraction layers can be confusing. Index, query engine, retriever, synthesizer—the terminology requires investment to master.","verdict":"The best data plumbing for LLM applications.","addedAt":"2026-08-18","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","active"],"capabilities":["rag","data"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":52560,"starsPerWeek":254,"access":"free","archived":false,"repo":{"createdAt":"2022-11-02","pushedAt":"2026-10-14","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/llamaindex","html":"/gems/llamaindex","markdown":"/gems/llamaindex.md"}},{"slug":"needle","name":"Needle 2","tagline":"14MB on-device tool-calling model","category":"agents","url":"https://github.com/cactus-compute/needle","github":"https://github.com/cactus-compute/needle","pricing":"open-source","hook":"Needle 2 is a 14MB model that runs tool-calling on-device.","whatsGood":"Cactus Compute built a tool-calling model that fits in 14MB. Runs entirely on-device—no API calls, no latency, no costs per call. Purpose-built for structured function calling, not general chat. The small footprint makes it embeddable anywhere.","theCatch":"Not a chat LLM. Needle does one thing: parse user intent into tool calls. You still need the tools and the orchestration around it. Do not expect conversation or reasoning.","verdict":"Tiny on-device tool-caller. One job, done locally.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","local-first","permissive","active"],"capabilities":["on-device","tool-calling"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":14087,"starsPerWeek":415,"access":"free","archived":false,"repo":{"createdAt":"2026-02-24","pushedAt":"2026-10-18","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/needle","html":"/gems/needle","markdown":"/gems/needle.md"}},{"slug":"mastra","name":"Mastra","tagline":"TypeScript agent runtime with typed agents and workflows","category":"agents","url":"https://mastra.ai","github":null,"pricing":"open-source","hook":"Mastra is the TypeScript runtime for building typed agents, workflows, and evals.","whatsGood":"Type-safe agent development in TypeScript. Define agents with proper types, compose them into workflows, run evals to measure quality. The framework approach means structure and reusability. Good fit for teams building agent systems rather than one-off scripts.","theCatch":"Framework, not a coding CLI. Mastra helps you build agents; it is not an agent itself. Learning curve for the runtime concepts. You are committing to their abstractions.","verdict":"TypeScript agent framework for teams building agent systems.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","evals","workflows"],"capabilities":["typescript","workflows"],"surfaces":["Self-host"],"ecosystem":["Node.js"],"licenseModel":"OSS/MIT","githubStars":null,"starsPerWeek":null,"access":"free","archived":false,"repo":null,"links":{"self":"/api/gems/mastra","html":"/gems/mastra","markdown":"/gems/mastra.md"}},{"slug":"hyperswitch","name":"Hyperswitch","tagline":"Apache-2.0 payment orchestrator you can self-host","category":"payments","url":"https://hyperswitch.io","github":"https://github.com/juspay/hyperswitch","pricing":"open-source","hook":"Hyperswitch is the payments switch you run when Stripe-as-the-whole-stack is the lock-in.","whatsGood":"Apache-2.0. Rust orchestrator: routing, vault, retries, 3DS, 90+ connectors. Community Edition is start-building-for-free on the pricing page. About page: free, open-source under Apache 2.0. This is Juspay's control center, open-sourced.","theCatch":"Community Edition means you run PCI scope. Cloud and Self-Host Enterprise are Contact Us. The 500+ global enterprises line is Juspay production experience, not a Hyperswitch-customer logo wall.","verdict":"Apache-2.0 payments orchestration. Self-host is free; PCI is not optional.","addedAt":"2026-09-03","featured":false,"domain":"Payments","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","payments","active"],"capabilities":["payments","self-host"],"surfaces":["Self-host"],"ecosystem":["Rust"],"licenseModel":"Apache-2.0","githubStars":45360,"starsPerWeek":217,"access":"free","archived":false,"repo":{"createdAt":"2022-10-17","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/hyperswitch","html":"/gems/hyperswitch","markdown":"/gems/hyperswitch.md"}},{"slug":"hatchet","name":"Hatchet","tagline":"MIT durable tasks/agents on Postgres; Cloud free tier real","category":"agents","url":"https://hatchet.run","github":"https://github.com/hatchet-dev/hatchet","pricing":"freemium","hook":"Hatchet is durable workflow/agent orchestration on the Postgres you already run when Temporal is the heavier cluster.","whatsGood":"MIT engine. Queues, retries, durable workflows, DAG-ish pipelines, Python/TS/Go/Ruby SDKs. Hatchet Lite single image for low throughput; Compose + RabbitMQ for prod. Cloud: first 100k task runs/mo free (Developer).","theCatch":"Vs Temporal/Inngest/Trigger.dev: Postgres-backed durable tasks + agent runs — not Temporal’s own cluster story, not Inngest/Trigger event-function DX alone. Cloud paid past free run allotment. Self-host is $0 software.","verdict":"MIT durable tasks on Postgres. Cloud free tier real; self-host forever.","addedAt":"2026-09-04","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["free-tier","permissive","postgres","workflows","active"],"capabilities":["workflows","durable","agents"],"surfaces":["Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":8139,"starsPerWeek":55,"access":"free","archived":false,"repo":{"createdAt":"2023-12-15","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/hatchet","html":"/gems/hatchet","markdown":"/gems/hatchet.md"}},{"slug":"smolagents","name":"smolagents","tagline":"Apache-2.0 minimal code-agent framework from Hugging Face","category":"agents","url":"https://huggingface.co/docs/smolagents","github":"https://github.com/huggingface/smolagents","pricing":"open-source","hook":"smolagents is the few-thousand-line agent kit when LangGraph is the heavy graph runtime and you want agents that write Python to call tools.","whatsGood":"Apache-2.0. CodeAgent (executable Python) + ToolCallingAgent (JSON). Hub-native tools, multi-agent orchestration, optional E2B sandbox. First-class HF model routing. Tiny core you can read.","theCatch":"Minimalist — not Temporal durable workflows or production multi-tenant agent platform. Code execution needs sandbox discipline. Vs CrewAI/PydanticAI: HF Hub gravity, less enterprise packaging. You own evals, memory, and deploy.","verdict":"Apache-2.0 code-first agents. Free library; sandbox/ops are yours.","addedAt":"2026-09-04","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","active"],"capabilities":["agents","code","tools"],"surfaces":["Self-host"],"ecosystem":["Python"],"licenseModel":"Apache-2.0","githubStars":29918,"starsPerWeek":306,"access":"free","archived":false,"repo":{"createdAt":"2024-12-05","pushedAt":"2026-10-06","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/smolagents","html":"/gems/smolagents","markdown":"/gems/smolagents.md"}},{"slug":"agno","name":"Agno","tagline":"Apache-2.0 Python agent framework + AgentOS runtime (ex-Phidata)","category":"agents","url":"https://www.agno.com","github":"https://github.com/agno-agi/agno","pricing":"open-source","hook":"Agno is the Phidata rebrand: a fast Python agent SDK with an AgentOS runtime you can actually serve.","whatsGood":"Apache-2.0. Build agents/teams/workflows in the Agno SDK, run them as a FastAPI AgentOS service (sessions, memory, tracing, optional MCP/A2A interfaces), manage via Control Plane. Model-agnostic toolkits; self-host the stack. Phidata is archived—this is the live line.","theCatch":"Vs LangGraph (graphs), CrewAI (roles), Mastra (TypeScript): Agno is Python performance + AgentOS productization, not a coding CLI like Claude Code/OpenCode. Phidata docs/imports are dead—use Agno. Control Plane and hosted conveniences are optional; day-one is OSS SDK+runtime you operate.","verdict":"Apache-2.0 Python agents + AgentOS. Ship on Agno, not Phidata nostalgia.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","active"],"capabilities":["agents","python","runtime"],"surfaces":["SDK","Self-host"],"ecosystem":["Python"],"licenseModel":"Apache-2.0","githubStars":42802,"starsPerWeek":184,"access":"free","archived":false,"repo":{"createdAt":"2022-05-04","pushedAt":"2026-10-17","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/agno","html":"/gems/agno","markdown":"/gems/agno.md"}},{"slug":"ag2","name":"AG2","tagline":"Apache-2.0 open-source AgentOS; AutoGen community continuation","category":"agents","url":"https://www.ag2.ai","github":"https://github.com/ag2ai/ag2","pricing":"open-source","hook":"AG2 is the community AutoGen continuation — build multi-agent systems under open governance, not a Microsoft seat.","whatsGood":"Apache-2.0 (ag2ai/ag2). Live brand for the AutoGen OSS line: pip install ag2, docs at docs.ag2.ai. v1 is protocol-driven (import ag2); Classic ConversableAgent/GroupChat lives in ag2-classic if you need the old API. Discord + open governance pitch.","theCatch":"Vs microsoft/autogen (maintenance; MS points new work to Microsoft Agent Framework) and LangGraph/CrewAI: AG2 is the community AutoGen fork, not Microsoft’s enterprise successor. AG2 v1 is not a drop-in from Classic — pin ag2-classic / import autogen if you need the old surface. One primary entry; do not twin with dead AutoGen marketing.","verdict":"Apache-2.0 AutoGen community AgentOS. Use AG2 for OSS; Microsoft Agent Framework if you want the MS successor.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","active"],"capabilities":["multi-agent","orchestration","python"],"surfaces":["Self-host"],"ecosystem":["Python"],"licenseModel":"Apache-2.0","githubStars":5003,"starsPerWeek":49,"access":"free","archived":false,"repo":{"createdAt":"2024-11-11","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/ag2","html":"/gems/ag2","markdown":"/gems/ag2.md"}},{"slug":"voltagent","name":"VoltAgent","tagline":"MIT TypeScript agent framework + optional VoltOps console","category":"agents","url":"https://voltagent.dev","github":"https://github.com/VoltAgent/voltagent","pricing":"open-source","hook":"VoltAgent is the TypeScript agent engineering platform: typed agents, workflows, MCP, and an optional VoltOps ops console.","whatsGood":"MIT (@voltagent/core). Memory, RAG, guardrails, tools, MCP, voice, supervisor/sub-agents, workflow engine. Deploy Node or serverless (Workers/edge). VoltOps Console (cloud or self-host) for observability/evals/deploy — not required to build.","theCatch":"Vs Mastra (TS runtime) and CrewAI/Agno (Python): VoltAgent is TS framework + platform surface, not a terminal coding agent. VoltOps is the paid/hosted ops layer; core stays MIT. Framework lock-in applies — you are building agents, not buying Claude Code.","verdict":"MIT TypeScript agent framework. VoltOps optional; core is the gem.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","sdk","self-host","permissive","workflows","active"],"capabilities":["typescript","workflows","mcp"],"surfaces":["MCP","Self-host"],"ecosystem":["Node.js"],"licenseModel":"OSS/MIT","githubStars":10830,"starsPerWeek":137,"access":"free","archived":false,"repo":{"createdAt":"2025-04-16","pushedAt":"2026-10-15","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/voltagent","html":"/gems/voltagent","markdown":"/gems/voltagent.md"}},{"slug":"rivet","name":"Rivet","tagline":"MIT visual AI agent graph IDE + TypeScript runtime (Ironclad)","category":"agents","url":"https://rivet.ironcladapp.com","github":"https://github.com/Ironclad/rivet","pricing":"open-source","hook":"Rivet (Ironclad) is the desktop IDE for drawing LLM agent graphs and embedding them in TypeScript apps.","whatsGood":"MIT. Electron IDE for prompt chains/agents plus @ironclad/rivet-core / rivet-node to run .rivet graphs in Node. External Call nodes bridge your app code. Distinct Ironclad product — not Rivet game engines or unrelated Rivets.","theCatch":"Vs LangGraph (code-first graphs) and Mastra/VoltAgent (TS frameworks): Rivet is visual-graph IDE + embed library, not a coding CLI and not LangChain. Graph quality still depends on your prompts/providers; External Calls need a host app + remote debugger. Desktop-first workflow.","verdict":"MIT Ironclad agent-graph IDE. Visual first; embed with rivet-node.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["desktop-app","self-host","permissive","rag","active"],"capabilities":["graphs","ide","typescript"],"surfaces":["Self-host"],"ecosystem":["Node.js","Browser"],"licenseModel":"OSS/MIT","githubStars":4709,"starsPerWeek":26,"access":"free","archived":false,"repo":{"createdAt":"2023-04-22","pushedAt":"2026-08-26","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/rivet","html":"/gems/rivet","markdown":"/gems/rivet.md"}},{"slug":"composio","name":"Composio","tagline":"Freemium agent toolkits + managed OAuth across 1000+ apps","category":"agents","url":"https://composio.dev","github":"https://github.com/ComposioHQ/composio","pricing":"freemium","hook":"Composio gives agents authenticated tools across 1000+ apps so you stop hand-rolling OAuth for every SaaS.","whatsGood":"OSS SDKs (TS/Python) + CLI on ComposioHQ/composio; hosted catalog with managed OAuth, sessions, tool search, and sandboxed execution. Wire Claude/Cursor/MCP or native tool-calling. Sessions discover/authorize/execute without stuffing hundreds of tool schemas into context.","theCatch":"Vs github-mcp (one official GitHub surface) and raw MCP servers: Composio is a multi-app auth/tool platform — you depend on their connector quality and hosted auth. Free tier exists; production is seat/usage freemium. Not a coding agent and not a browser fleet.","verdict":"Freemium agent tool/auth fabric. Great connectors; you rent the OAuth middle.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","free-tier","permissive","auth","active"],"capabilities":["tools","auth","mcp"],"surfaces":["MCP","Hosted"],"ecosystem":["Python","Node.js","Multi-platform"],"licenseModel":"Freemium","githubStars":30635,"starsPerWeek":221,"access":"free","archived":false,"repo":{"createdAt":"2024-02-23","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/composio","html":"/gems/composio","markdown":"/gems/composio.md"}},{"slug":"google-adk","name":"Google ADK","tagline":"Apache-2.0 code-first agent build/eval/deploy kit (Google)","category":"agents","url":"https://adk.dev","github":"https://github.com/google/adk-python","pricing":"open-source","hook":"Google ADK is the code-first toolkit to build, evaluate, and deploy agents — software-dev shaped, not prompt-paste.","whatsGood":"Apache-2.0. Primary Python line (`pip install google-adk`, google/adk-python); docs at adk.dev. Multi-agent orchestration, tools, eval, local run and deploy targets. Also TS/Go/Java siblings. Model-agnostic enough to run off-GCP; Gemini/Vertex paths are first-class when you want them.","theCatch":"Vs Gemini CLI / Antigravity CLI (terminal coding agents), CrewAI/LangGraph/Agno (orchestration frameworks), and Daytona/E2B (sandboxes): ADK is the Google build/eval/deploy kit, not a coding CLI and not a sandbox fleet. Vertex/GCP gravity is real for enterprise deploy; OSS still runs without a Google Cloud bill if you stay local/BYO.","verdict":"Apache-2.0 Google agent kit. Code-first build/eval/deploy; Vertex optional, not mandatory.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","evals","active"],"capabilities":["agents","eval","deploy"],"surfaces":["Self-host"],"ecosystem":["Python","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":21906,"starsPerWeek":271,"access":"free","archived":false,"repo":{"createdAt":"2025-04-01","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/google-adk","html":"/gems/google-adk","markdown":"/gems/google-adk.md"}},{"slug":"dspy","name":"DSPy","tagline":"MIT program LMs with signatures and optimizers (Stanford NLP)","category":"agents","url":"https://dspy.ai","github":"https://github.com/stanfordnlp/dspy","pricing":"open-source","hook":"DSPy programs language models — signatures and optimizers instead of brittle prompt strings.","whatsGood":"MIT (stanfordnlp/dspy). Declarative Python modules, compilers/optimizers that tune prompts and weights, community + Stanford NLP lineage. Docs at dspy.ai. Fits classifiers, RAG, and agent loops when you want systematic improvement over hand-tuned templates.","theCatch":"Vs LangChain/LangGraph (orchestration graphs) and Promptfoo (YAML eval/red-team): DSPy is program-and-optimize LMs, not an MCP-first runtime and not a hosted eval SaaS. Optimizer runs cost tokens/compute; you still bring models and data. Not a drop-in coding agent.","verdict":"MIT LM programming framework. Optimize modules; do not confuse it with MCP runtime glue.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","active"],"capabilities":["optimizers","prompts","python"],"surfaces":["Self-host"],"ecosystem":["Python"],"licenseModel":"OSS/MIT","githubStars":38687,"starsPerWeek":196,"access":"free","archived":false,"repo":{"createdAt":"2023-01-09","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/dspy","html":"/gems/dspy","markdown":"/gems/dspy.md"}},{"slug":"camel","name":"CAMEL","tagline":"Apache-2.0 multi-agent framework — scaling laws of agents","category":"agents","url":"https://www.camel-ai.org","github":"https://github.com/camel-ai/camel","pricing":"open-source","hook":"CAMEL is the research-born multi-agent framework chasing scaling laws of agents — roles, societies, and simulated environments.","whatsGood":"Apache-2.0 (camel-ai/camel). Multi-agent roles, tool use, environments, and large-scale agent society experiments. Docs at docs.camel-ai.org / camel-ai.org. Active research + builder community; pip-installable Python stack.","theCatch":"Vs CrewAI / AG2 (product multi-agent orchestration) and OpenHands / SWE-agent (coding agents): CAMEL is multi-agent research/framework breadth, not a turnkey SWE seat. Eigent is a separate commercial cousin — do not treat CAMEL OSS as Eigent cloud. You own model spend and orchestration choices.","verdict":"Apache-2.0 multi-agent framework. Scale agent societies; Eigent is a different product.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","active"],"capabilities":["multi-agent","research","python"],"surfaces":["Self-host"],"ecosystem":["Python"],"licenseModel":"Apache-2.0","githubStars":17932,"starsPerWeek":96,"access":"free","archived":false,"repo":{"createdAt":"2023-03-17","pushedAt":"2026-10-12","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/camel","html":"/gems/camel","markdown":"/gems/camel.md"}},{"slug":"genkit","name":"Genkit","tagline":"Apache-2.0 Google/Firebase agentic-app flows — JS, Go, Python","category":"agents","url":"https://genkit.dev","github":"https://github.com/genkit-ai/genkit","pricing":"open-source","hook":"Genkit is Google/Firebase’s open-source framework for full-stack AI and agentic app flows — prompts, tools, and deploy helpers.","whatsGood":"Apache-2.0 (firebase/genkit). SDKs for JavaScript/TypeScript, Go, and Python. Structured output, tool calling, flows, local Developer UI/CLI. Providers include Google GenAI, OpenAI, Ollama, and more. Ship to Firebase, Cloud Run, or your own host. Docs at genkit.dev.","theCatch":"Vs Vercel AI SDK (TS app AI primitives) and Google ADK (agent build/eval/deploy kit): Genkit is Firebase-shaped full-stack AI/agent flows, not a multi-agent research OS and not a coding CLI. Firebase/GCP paths may bill when you leave local — OSS core still runs without Google Cloud if you stay self-hosted/BYO providers.","verdict":"Apache-2.0 Firebase/Google AI app framework. Agentic flows in code; watch GCP meters in prod.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","active"],"capabilities":["flows","firebase","sdk"],"surfaces":["SDK","Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":6529,"starsPerWeek":51,"access":"free","archived":false,"repo":{"createdAt":"2024-04-29","pushedAt":"2026-10-18","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/genkit","html":"/gems/genkit","markdown":"/gems/genkit.md"}},{"slug":"bifrost","name":"Bifrost","tagline":"Apache-2.0 AI gateway — perf angle vs LiteLLM (Maxim)","category":"agents","url":"https://docs.getbifrost.ai","github":"https://github.com/maximhq/bifrost","pricing":"freemium","hook":"Bifrost is Maxim’s Apache-2.0 AI gateway — OpenAI-compatible multi-provider proxy with a hard performance pitch versus LiteLLM.","whatsGood":"Apache-2.0 (maximhq/bifrost). `npx -y @maximhq/bifrost` zero-config gateway; Go SDK path too. Fallbacks, load balancing, MCP tools, semantic caching, governance/virtual keys. Docs at docs.getbifrost.ai. Benchmarks claim large overhead wins vs LiteLLM at high RPS — verify on your hardware.","theCatch":"Vs LiteLLM: same gateway job with a Go/perf positioning — not a model marketplace. Maxim upsell: enterprise/clustering/adaptive LB/OIDC provisioning and getmaxim.ai Bifrost enterprise pages are the commercial gravity; OSS gateway still runs locally. You still pay upstream model providers.","verdict":"Apache-2.0 AI gateway. Fast LiteLLM alternative; Maxim enterprise is the loud paid path.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["free-tier","permissive","active"],"capabilities":["llm","proxy","gateway"],"surfaces":["Hosted"],"ecosystem":["Go","Multi-platform"],"licenseModel":"Freemium","githubStars":8924,"starsPerWeek":108,"access":"free","archived":false,"repo":{"createdAt":"2025-03-19","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/bifrost","html":"/gems/bifrost","markdown":"/gems/bifrost.md"}},{"slug":"deskcommcrm","name":"DeskcommCRM","tagline":"MIT self-hosted AI sales OS/CRM — MCP-ready","category":"agents","url":"https://www.deskcomm.com.br/","github":"https://github.com/melgarafael/DeskcommCRM","pricing":"open-source","hook":"DeskcommCRM is a self-hosted AI sales OS — CRM with native agents and WhatsApp (WAHA), pitched as an open alternative to Intercom-class chat sales stacks.","whatsGood":"MIT (melgarafael/DeskcommCRM). Self-hosted multi-tenant CRM, AI agents, WhatsApp pipeline, LGPD-minded positioning, MCP-ready API surface. Docker/VPS install path; site at deskcomm.com.br.","theCatch":"Vs Chatwoot (open inbox/helpdesk) and Composio (agent tool auth/connectors): DeskcommCRM is an AI sales OS/CRM with WhatsApp agents, not a general shared inbox alone and not a universal tool marketplace. HostGator affiliate/setup kit is infra convenience, not product lock — you can host elsewhere.","verdict":"MIT self-hosted AI sales CRM. Agents+WhatsApp vs Chatwoot/Composio; HostGator kit ≠ lock.","addedAt":"2026-09-16","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","self-host","permissive","active","young-repo"],"capabilities":["crm","agents","mcp"],"surfaces":["MCP","Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":4693,"starsPerWeek":188,"access":"free","archived":false,"repo":{"createdAt":"2026-04-28","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/deskcommcrm","html":"/gems/deskcommcrm","markdown":"/gems/deskcommcrm.md"}},{"slug":"thurbox","name":"Thurbox","tagline":"MIT tmux TUI/CLI for local AI agent orchestration","category":"agents","url":"https://thurbox.thurbeen.eu/","github":"https://github.com/Thurbeen/thurbox","pricing":"open-source","hook":"Thurbox is a tmux-backed TUI that keeps multiple coding-agent CLIs alive in parallel — local orchestration without a cloud harness.","whatsGood":"MIT (Thurbeen/thurbox). `thurbox` TUI + `thurbox-cli` headless; sessions persist in tmux so agents keep working when you detach. Repo picker, worktree mode, multi-agent panes. Docs/install at thurbox.thurbeen.eu (Show HN-shaped local harness).","theCatch":"Vs Atlas (pacifio agent source-control) and DeskcommCRM (self-hosted AI sales CRM): Thurbox is a local tmux/TUI orchestrator for coding-agent CLIs — not session+commit memory like Atlas, not a CRM. Early stars (~67★); needs tmux ≥3.2 (or psmux on Windows) plus at least one agent CLI you already pay for or self-host.","verdict":"MIT local agent TUI. tmux parallel agents vs Atlas/DeskcommCRM; young but concrete.","addedAt":"2026-09-16","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["cli","tui","self-host","permissive","coding-agents","workflows","active"],"capabilities":["cli"],"surfaces":["CLI","Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":93,"starsPerWeek":3,"access":"free","archived":false,"repo":{"createdAt":"2026-02-12","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/thurbox","html":"/gems/thurbox","markdown":"/gems/thurbox.md"}},{"slug":"weknora","name":"WeKnora","tagline":"MIT self-host RAG/knowledge platform (Tencent)","category":"agents","url":"https://weknora.weixin.qq.com","github":"https://github.com/Tencent/WeKnora","pricing":"open-source","hook":"WeKnora is Tencent’s open-source LLM knowledge platform — turn documents into queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.","whatsGood":"MIT (Tencent/WeKnora) with bundled third-party notices. Self-host RAG/knowledge: ingest docs, retrieve, agent reasoning, wiki-shaped maintenance. Go-heavy stack, Docker paths, multi-tenant/knowledge-base positioning. Large adoption (~26k★). Forever OSS path is the GitHub tree you run yourself.","theCatch":"Optional weknora.weixin.qq.com host is a separate product surface from the MIT self-host forever path — do not treat Tencent’s hosted WeChat-adjacent portal as the license or the ops model. LICENSE ships with THIRD_PARTY_NOTICES / licenses/ for bundled deps — MIT on the project, not a free pass on every vendored component. Vs LibreChat / AnythingLLM: WeKnora is knowledge/RAG+agent+wiki platform weight, not only a multi-provider chat UI.","verdict":"MIT self-host knowledge/RAG. OSS forever ≠ weknora.weixin.qq.com host; third-party notices are loud.","addedAt":"2026-09-17","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","license-unclear","rag","active","fast-stars"],"capabilities":["rag","knowledge","self-host"],"surfaces":["Self-host"],"ecosystem":["Go","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":34646,"starsPerWeek":534,"access":"free","archived":false,"repo":{"createdAt":"2025-07-22","pushedAt":"2026-10-19","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/weknora","html":"/gems/weknora","markdown":"/gems/weknora.md"}},{"slug":"ecc","name":"ECC","tagline":"MIT agent ecosystem / coordination surface","category":"agents","url":"https://ecc.tools","github":"https://github.com/affaan-m/ECC","pricing":"open-source","hook":"ECC is an agent-harness performance layer — skills, instincts, memory, security, and research-first workflows across Claude Code, Codex, OpenCode, Cursor, and friends.","whatsGood":"MIT (affaan-m/ECC). Ecosystem/coordination surface for coding agents: skills packs, memory/instincts, security helpers, research-first development patterns. Site at ecc.tools; viral adoption (~260k★). Aimed at teams already living inside agent CLIs who want a shared optimization layer.","theCatch":"Still depends on host agents (Claude Code / Codex / Cursor / etc.) — ECC coordinates and tunes the harness; it is not a replacement coding agent. Vs OpenSpec (spec-driven change tooling) and Thurbox (tmux multi-agent TUI): ECC is the skills/memory/security ecosystem pack, not a spec framework and not a pane orchestrator.","verdict":"MIT agent coordination ecosystem. Skills+memory+security on top of host agents — not a host itself.","addedAt":"2026-09-17","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","coding-agents","skills","security","workflows","active","fast-stars"],"capabilities":["agents","skills","memory"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":283433,"starsPerWeek":7226,"access":"free","archived":false,"repo":{"createdAt":"2026-01-18","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/ecc","html":"/gems/ecc","markdown":"/gems/ecc.md"}},{"slug":"octop","name":"Octop","tagline":"MIT self-hosted multi-user multi-agent platform","category":"agents","url":"https://github.com/TencentCloud/Octop","github":"https://github.com/TencentCloud/Octop","pricing":"open-source","hook":"Octop is a self-hosted control room for teams running multiple agents, not just another single-user chat shell.","whatsGood":"MIT (TencentCloud/Octop). Multi-user, multi-agent self-hosting with MCP and ACP connectivity, RAG, and IM-shaped collaboration surfaces. It gives teams one place to operate agent workloads while keeping the deployment and model choices on their side.","theCatch":"Vs LibreChat, OpenClaw, and nanobot: Octop is the team/multi-agent platform surface, not merely a multi-provider chat UI, a personal agent gateway, or a lightweight bot runtime. The self-host bargain moves auth, model keys, storage, upgrades, and operational boundaries onto you.","verdict":"MIT multi-user multi-agent platform. MCP/ACP/RAG/IM in one self-hosted surface; you own the control plane.","addedAt":"2026-09-18","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","self-host","permissive","active","young-repo","fast-stars"],"capabilities":["multi-agent","mcp","self-host"],"surfaces":["MCP","Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":9951,"starsPerWeek":672,"access":"free","archived":false,"repo":{"createdAt":"2026-07-08","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/octop","html":"/gems/octop","markdown":"/gems/octop.md"}},{"slug":"agentbox","name":"AgentBox","tagline":"Teleport a repo into parallel sandboxed agent VMs","category":"agents","url":"https://agent-box.sh","github":"https://github.com/madarco/agentbox","pricing":"open-source","hook":"AgentBox teleports your project into a dedicated VM so multiple agents can run in parallel without wrecking your laptop worktree.","whatsGood":"MIT (madarco/agentbox). `agentbox claude` (and friends) spin sandboxed VMs locally or in the cloud with your project, tools, and sync/git flows. Docs at agent-box.sh cover teleport, services, E2B/Daytona/Hetzner/Vercel/DO backends, and terminal multiplexers. npm `@madarco/agentbox`. ★~429.","theCatch":"Vs E2B / Daytona (ephemeral sandboxes) and Coder (durable remote workspaces): AgentBox is the one-command teleport + multi-agent box launcher, not a full workspace control plane. Docker/OrbStack-class virtualization and cloud backend credentials are on you; cloud minutes are not “free forever.”","verdict":"MIT parallel agent VM launcher. Reach for it when agents need isolated boxes, not another shared worktree.","addedAt":"2026-09-18","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["self-host","permissive","sandbox","git","active","young-repo"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":538,"starsPerWeek":23,"access":"free","archived":false,"repo":{"createdAt":"2026-05-12","pushedAt":"2026-10-07","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/agentbox","html":"/gems/agentbox","markdown":"/gems/agentbox.md"}},{"slug":"onecli","name":"OneCLI","tagline":"Apache-2.0 team agent harness — sandboxed personal agents + credential gateway","category":"agents","url":"https://onecli.sh","github":"https://github.com/onecli/onecli","pricing":"open-source","hook":"OneCLI is the open-source harness for giving every teammate a sandboxed personal agent — credentials inject at a gateway so keys never live in the agent prompt.","whatsGood":"Apache-2.0 (onecli/onecli). Self-host with pnpm; cloud at onecli.sh. Grew from a Rust credential vault into a multiplayer agent platform: per-person agents, sandbox, policy, and secret injection for stacks like Hermes / OpenClaw-class runners. The wedge is team ops — replicate one good personal agent across a company without copying API keys into every session.","theCatch":"Vs AgentBox / Daytona / E2B: OneCLI is the team harness + credential gateway story, not a generic cloud sandbox product. Vs Infisical: secrets are a means; the product is running agents safely for many people. You still bring model spend and decide which agent CLIs to wire. Cloud convenience is optional — OSS self-host is the forever path.","verdict":"OSS team agent harness with credential injection. Sandbox + gateway for many personal agents; not just another secrets vault.","addedAt":"2026-09-19","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["self-host","permissive","sandbox","active"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":3580,"starsPerWeek":111,"access":"free","archived":false,"repo":{"createdAt":"2026-03-08","pushedAt":"2026-10-14","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/onecli","html":"/gems/onecli","markdown":"/gems/onecli.md"}},{"slug":"flue","name":"Flue","tagline":"Astro TypeScript harness for autonomous sandbox agents","category":"agents","url":"https://github.com/withastro/flue","github":"https://github.com/withastro/flue","pricing":"open-source","hook":"Flue is Astro's agent harness framework: the agent is a TypeScript function composed with model, sandbox, skills, and tools instead of another chat SDK.","whatsGood":"Apache-2.0 (withastro/flue). Agents are programmable TypeScript harnesses (`useModel`, `useSandbox`, `useSkill`, `useTool`) with local or container sandboxes, sessions, and filesystem access. It targets the Claude Code / Codex style of autonomous work rather than scripted chatbot chains, and can run via CLI or a hosted runtime you choose.","theCatch":"Vs AgentBox and Octop: Flue is a TypeScript framework for authoring agent harnesses, not a multi-user agent platform or a generic sandbox/VM launcher. You still bring models, credentials, and operational judgment; a dead flue.dev marketing domain is not required — the GitHub repo is the source of truth.","verdict":"Apache-2.0 Astro agent harness. Use when you want programmable TypeScript agents with real sandboxes; bring your own models and host.","addedAt":"2026-09-20","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","sandbox","skills","active"],"capabilities":["agents","sandbox","typescript"],"surfaces":["SDK","Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":8489,"starsPerWeek":233,"access":"free","archived":false,"repo":{"createdAt":"2026-02-07","pushedAt":"2026-10-17","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/flue","html":"/gems/flue","markdown":"/gems/flue.md"}},{"slug":"substrate","name":"Agent Substrate","tagline":"High-density secure sandbox runtime for idle-heavy agent workloads","category":"agents","url":"https://github.com/agent-substrate/substrate","github":"https://github.com/agent-substrate/substrate","pricing":"open-source","hook":"Agent Substrate runs many agent sandboxes densely by multiplexing idle actors onto fewer workers with fast suspend/resume and kernel/network isolation.","whatsGood":"Apache-2.0 (agent-substrate/substrate). Purpose-built execution runtime for agent-like workloads: microVM and gVisor sandboxes, sub-500ms resume claims, zero-trust isolation, and Kubernetes-based worker lifecycle with agent-aware scheduling. It is infrastructure for running agents at scale, not another agent SDK.","theCatch":"Vs AgentBox and Flue: Substrate is the dense sandbox/runtime control plane on Kubernetes, not a developer ADE and not a TypeScript harness framework. Ops cost is real — you need cluster competence. README notes it is not an officially supported Google product.","verdict":"Apache-2.0 agent sandbox runtime. Reach for dense suspend/resume isolation on K8s; bring cluster ops.","addedAt":"2026-09-20","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","sandbox","active","young-repo"],"capabilities":["sandbox","kubernetes","runtime"],"surfaces":["Self-host"],"ecosystem":["Go","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":5530,"starsPerWeek":243,"access":"free","archived":false,"repo":{"createdAt":"2026-05-13","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/substrate","html":"/gems/substrate","markdown":"/gems/substrate.md"}},{"slug":"ax-agent-executor","name":"AX (Agent Executor)","tagline":"Apache-2.0 Go CLI and Kubernetes CRDs for running agent tasks","category":"agents","url":"https://agentexecutor.io","github":"https://github.com/google/ax","pricing":"open-source","hook":"AX (Agent Executor) turns agent work into Kubernetes-native Tasks, Workspaces, and Gateways that can run on Agent Substrate instead of disappearing inside a chat session.","whatsGood":"Apache-2.0. The Go CLI and Kubernetes CRDs give an agent run a durable execution shape: a Task for the work, a Workspace for its files, and a Gateway for the services around it. That makes AX useful when the problem is operating agent workloads on Kubernetes, not merely calling a model from an SDK.","theCatch":"This is AX Agent Executor, not the live AX Check (Gauge) evaluation product — the names collide, so use the ax-agent-executor slug. Vs Agent Substrate: AX is the task/CLI/CRD experience on top of an agent runtime; vs OneCLI and Dexto: it is Kubernetes execution infrastructure, not a general desktop agent; vs Cua: it is not a browser/computer-use surface. Kubernetes and Agent Substrate operations are still your problem.","verdict":"Apache-2.0 agent executor for Kubernetes. Use the ax-agent-executor CLI and CRDs when durable Task/Workspace/Gateway primitives matter; do not confuse it with AX Check.","addedAt":"2026-09-21","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["cli","self-host","permissive","kubernetes","active"],"capabilities":["cli","kubernetes","runtime"],"surfaces":["CLI","Self-host"],"ecosystem":["Go","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":13944,"starsPerWeek":479,"access":"free","archived":false,"repo":{"createdAt":"2026-03-30","pushedAt":"2026-10-18","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/ax-agent-executor","html":"/gems/ax-agent-executor","markdown":"/gems/ax-agent-executor.md"}},{"slug":"unreal-agent","name":"Unreal Agent","tagline":"Async-first agent harness with durable sessions and operations","category":"agents","url":"https://unreallabs.ai","github":"https://github.com/unreallabsai/unreal-agent","pricing":"open-source","hook":"Unreal Agent is an async-first harness: session inbox idempotency, append-only history, tool translators that emit serializable operations, and a swappable operation manager.","whatsGood":"MIT (unreallabsai/unreal-agent). Clear component split — coordinator, session store, context builder, LLM adapter, tool registry — with versioned serializable sessions/operations and fork/recovery paths. Docs/site at unreallabs.ai. Built for hosts that need durable tool work, not another chat REPL.","theCatch":"Vs Pi, Agentao, and OpenHands: this is a library harness with durable ops semantics, not a polished coding CLI, governed Python embed, or full autonomous developer product. You assemble adapters and storage; the README is architecture-first, not a one-command demo.","verdict":"MIT async harness when session durability and serializable tool ops matter more than a turnkey coding TUI.","addedAt":"2026-09-23","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["self-host","permissive","workflows","active","young-repo","fast-stars"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":2206,"starsPerWeek":540,"access":"free","archived":false,"repo":{"createdAt":"2026-09-21","pushedAt":"2026-10-06","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/unreal-agent","html":"/gems/unreal-agent","markdown":"/gems/unreal-agent.md"}},{"slug":"deerflow","name":"DeerFlow","tagline":"MIT long-horizon SuperAgent harness — sandboxes, memory, skills, gateway","category":"agents","url":"https://deerflow.tech","github":"https://github.com/bytedance/deer-flow","pricing":"open-source","hook":"DeerFlow 2.0 is ByteDance’s open SuperAgent harness: sub-agents, memory, sandboxes, extensible skills, and a gateway so long-horizon work is orchestrated rather than one-shot chatted.","whatsGood":"MIT (bytedance/deer-flow). Ground-up 2.0 rewrite (v1 deep-research stays on a 1.x branch). Docker Compose path, MCP/IM channels, tracing hooks, and multi-provider model config. Official demos at deerflow.tech; Python 3.12+ / Node 22+ stack.","theCatch":"Vs OpenHands, LangGraph, and Mastra: DeerFlow is a full SuperAgent product harness with sandboxes/memory/gateway, not an open coding agent alone, a graph orchestration library, or a TypeScript workflow framework. BYO model API keys; Docker sizing is resource-heavy (multi-vCPU/GB guidance); star velocity is extreme (~83k).","verdict":"MIT SuperAgent harness when you want sandboxes + skills + gateway in one OSS tree and will bring models plus Docker headroom.","addedAt":"2026-09-24","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","sandbox","skills","active","fast-stars"],"capabilities":["agents","sandbox","memory"],"surfaces":["Self-host"],"ecosystem":["Python","Node.js","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":84096,"starsPerWeek":1109,"access":"free","archived":false,"repo":{"createdAt":"2025-05-07","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/deerflow","html":"/gems/deerflow","markdown":"/gems/deerflow.md"}},{"slug":"cubesandbox","name":"CubeSandbox","tagline":"Apache concurrent lightweight agent sandbox — RustVMM/KVM, E2B-compatible","category":"agents","url":"https://github.com/TencentCloud/CubeSandbox","github":"https://github.com/TencentCloud/CubeSandbox","pricing":"open-source","hook":"CubeSandbox is Tencent Cloud’s high-density agent sandbox: RustVMM + KVM hardware isolation, sub-60ms create, E2B-compatible API, single-node or multi-node clusters.","whatsGood":"Apache-2.0 per LICENSE text (TencentCloud/CubeSandbox) with PyPI client and CNCF landscape listing. Hardware-level isolation, pause/resume, K8s deploy path, volume plugins, and template aliases. Aimed at concurrent agent workloads, not a single toy container.","theCatch":"Vs E2B, Daytona, and Cloudflare Sandbox: this is a self-hosted RustVMM/KVM sandbox control plane, not a hosted sandbox SaaS, a workspace platform, or an edge isolate runtime. GitHub license detect shows NOASSERTION even though LICENSE is Apache-2.0 — verify the tree before compliance sign-off (a detection caveat). Linux + KVM on x86_64 (ARM64 maturing); ops weight is real.","verdict":"Apache KVM sandbox when you need dense, E2B-shaped isolation on your own Linux hosts; confirm license text and hardware before betting the fleet.","addedAt":"2026-09-24","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["api","self-host","permissive","license-unclear","sandbox","active"],"capabilities":["sandbox","kvm","agents"],"surfaces":["API","Self-host"],"ecosystem":["Rust","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":12976,"starsPerWeek":472,"access":"free","archived":false,"repo":{"createdAt":"2026-04-10","pushedAt":"2026-10-19","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/cubesandbox","html":"/gems/cubesandbox","markdown":"/gems/cubesandbox.md"}},{"slug":"deepseek-harness","name":"DeepSeek Harness","tagline":"MIT everything-is-a-plugin agent harness (dsh) — Cordis, Web UI","category":"agents","url":"https://deepseek-harness.github.io/deepseek-harness/","github":"https://github.com/deepseek-ai/deepseek-harness","pricing":"open-source","hook":"DeepSeek Harness (`dsh`) is DeepSeek’s open agent harness: an everything-is-a-plugin architecture on Cordis, with a local Web UI via `npx @deepseek-ai/dsh web`.","whatsGood":"MIT (deepseek-ai/deepseek-harness). Plugin topic `dsh-plugin`, AGENTS.md for agent contributors, docs at deepseek-harness.github.io, Discord + Discussions. Run path is npm (`npx @deepseek-ai/dsh web` → localhost:3080) or pnpm from source. Not the similarly named Harness SDK: this is DeepSeek’s plugin harness product.","theCatch":"Vs DeerFlow and OpenHands: this is a Cordis plugin-harness + Web UI from DeepSeek, not a ByteDance SuperAgent product tree or an open coding agent alone. Developer preview with explicit COMPATIBILITY-BREAKING CHANGES; SAFETY.md says no security audit, model/plugin code can damage the host — prefer a disposable VM/container. Extreme star velocity (~235k aria-label) is discovery noise, not proof.","verdict":"MIT plugin agent harness when you want Cordis-shaped extensibility and will treat preview APIs plus SAFETY.md as real constraints.","addedAt":"2026-09-24","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["web-ui","self-host","permissive","active","young-repo","fast-stars"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":253471,"starsPerWeek":26254,"access":"free","archived":false,"repo":{"createdAt":"2026-08-13","pushedAt":"2026-10-17","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/deepseek-harness","html":"/gems/deepseek-harness","markdown":"/gems/deepseek-harness.md"}},{"slug":"rakazo","name":"Rakazo","tagline":"Apache OSS persistent AI teammates — BYO model and sandbox","category":"agents","url":"https://rakazo.com","github":"https://github.com/elie222/rakazo","pricing":"open-source","hook":"Rakazo is an open-source platform for persistent AI teammates — web, Electron desktop, and Expo mobile — with BYO model credentials and Docker/E2B/Daytona/Box computers.","whatsGood":"Apache-2.0 (elie222/rakazo). Bots keep conversations, memory, routines; voice via BYO ElevenLabs/OpenAI/Cartesia/Fish; browser/terminal/desktop access; peer delegation; Composio/Pipedream plus MCP/OpenAPI/Treg integrations. Published Compose images (`install-images.sh`) or pnpm source path; docs at rakazo.com / self-host guide.","theCatch":"Vs OpenClaw and DeerFlow: Rakazo is a self-host teammate product (web/desktop/mobile + sandboxes), not a multi-channel messaging agent alone or a SuperAgent research harness. Default install is curl|bash of install-images.sh; beta; BYO model + optional paid sandbox providers; Treg has its own license rider if you embed it.","verdict":"Apache self-host teammate stack when you want persistent bots you own with pluggable models/sandboxes; budget Docker and keys, not just the UI.","addedAt":"2026-09-24","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["desktop-app","self-host","permissive","sandbox","containers","active","young-repo"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":4623,"starsPerWeek":479,"access":"free","archived":false,"repo":{"createdAt":"2026-08-13","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/rakazo","html":"/gems/rakazo","markdown":"/gems/rakazo.md"}},{"slug":"aio-sandbox","name":"AIO Sandbox","tagline":"Apache one container: browser+shell+files+VS Code+Jupyter+MCP","category":"agents","url":"https://sandbox.agent-infra.com/","github":"https://github.com/agent-infra/sandbox","pricing":"open-source","hook":"AIO Sandbox packs browser, shell, files, VS Code, Jupyter, and MCP onto one shared-filesystem container so an agent gets a full workstation in a single image.","whatsGood":"Apache-2.0 (agent-infra/sandbox). One `docker run` path (ghcr.io/agent-infra/sandbox), PyPI/npm clients, VNC browser + code-server + MCP on the same FS. Docs and API at sandbox.agent-infra.com. Designed as self-host all-in-one, not a managed microVM product.","theCatch":"Vs E2B, Daytona, and CubeSandbox (Thu managed/microVM sandbox lane): this is a self-host all-in-one container workstation, not hosted Firecracker sandboxes, a managed workspace SaaS, or a KVM density control plane. Quickstart docs use `--security-opt seccomp=unconfined`; set `SANDBOX_API_KEY` or API/Jupyter/VNC stay open by default.","verdict":"Apache self-host all-in-one sandbox when one shared FS with browser+IDE+MCP beats a managed microVM — lock the API key and read the seccomp note before exposing it.","addedAt":"2026-09-25","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","editor-plugin","self-host","permissive","sandbox","containers","active"],"capabilities":["sandbox","browser","mcp"],"surfaces":["MCP","Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":6184,"starsPerWeek":98,"access":"free","archived":false,"repo":{"createdAt":"2025-08-06","pushedAt":"2026-09-14","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/aio-sandbox","html":"/gems/aio-sandbox","markdown":"/gems/aio-sandbox.md"}},{"slug":"agent-mcp","name":"Agent-MCP","tagline":"AGPL MCP-native multi-agent + shared knowledge graph","category":"agents","url":"https://github.com/rinadelph/Agent-MCP","github":"https://github.com/rinadelph/Agent-MCP","pricing":"open-source","hook":"Agent-MCP is an MCP-native multi-agent collaboration stack: specialized agents share a persistent knowledge graph, task board, and live visualization instead of one overflowing chat.","whatsGood":"AGPL-3.0 per LICENSE/README (rinadelph/Agent-MCP). Parallel agents, searchable memory bank, dashboard/fleet views, Python server path with MCP tooling. Aimed at coordinated software work where context must survive across agents and sessions.","theCatch":"Vs CrewAI and AG2: this is MCP-native multi-agent + shared graph for coding workflows, not a general role-orchestration framework alone. LICENSE/README are AGPL-3.0 while pyproject (and node metadata) still claim MIT — metadata conflict; treat AGPL as authoritative until packaging is fixed. Last push 2026-03-28; typical path expects an OpenAI key.","verdict":"AGPL MCP multi-agent graph when you need shared project memory across agents and can live with copyleft plus stale packaging metadata.","addedAt":"2026-09-25","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","self-host","copyleft","stale"],"capabilities":["mcp","multi-agent","graph"],"surfaces":["MCP","Self-host"],"ecosystem":["Python","Multi-platform"],"licenseModel":"AGPL-3.0","githubStars":1304,"starsPerWeek":17,"access":"free","archived":false,"repo":{"createdAt":"2025-05-13","pushedAt":"2026-03-28","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/agent-mcp","html":"/gems/agent-mcp","markdown":"/gems/agent-mcp.md"}},{"slug":"paperclip","name":"Paperclip","tagline":"MIT org control-plane for agent budgets, Board governance, heartbeats","category":"agents","url":"https://paperclip.ing","github":"https://github.com/paperclipai/paperclip","pricing":"open-source","hook":"Paperclip is the open-source company OS for agents at work: budgets, Board-style governance, and heartbeats so org control sits above the coding CLIs.","whatsGood":"MIT (paperclipai/paperclip). Free self-host via `npx paperclipai onboard --yes`, install.sh, or pnpm. Org control-plane lane — budgets + Board governance + heartbeats — when you need company-level agent ops, not another issue board. Homepage paperclip.ing; ★~85k on GitHub.","theCatch":"Vs Multica: Multica is a coding-agent issue board; Paperclip is company OS / org control. Vs OpenClaw: messaging employee, not budgeted Board governance. ★85k, but repo created 2026-03 — maturity may not match stars; piped install.sh is a supply-chain note, not a free pass.","verdict":"MIT org control-plane when agent budgets and Board governance matter more than another task board — verify install path and star-vs-age risk.","addedAt":"2026-09-26","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","coding-agents","active","fast-stars"],"capabilities":["governance","budgets","agents"],"surfaces":["Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":104253,"starsPerWeek":3151,"access":"free","archived":false,"repo":{"createdAt":"2026-03-02","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/paperclip","html":"/gems/paperclip","markdown":"/gems/paperclip.md"}},{"slug":"ruflo","name":"Ruflo","tagline":"MIT meta-harness / execution swarm on coding CLIs (ex Claude Flow)","category":"agents","url":"https://github.com/ruvnet/ruflo","github":"https://github.com/ruvnet/ruflo","pricing":"open-source","hook":"Ruflo is the MIT meta-harness that runs execution swarms inside coding CLIs — the rename lineage of ruvnet/claude-flow (ex Claude Flow).","whatsGood":"MIT (ruvnet/ruflo). `npx ruflo@latest init` plus free in-repo plugins. Coordinates multi-agent workflows on Claude Code / Codex / Hermes-class tools with adaptive memory and federation hooks. Use when you want a harness swarm *inside* the tools you already run.","theCatch":"Vs Pi / OpenCode / DeepSeek Harness: those are coding CLI/harness products; Ruflo is a meta-harness on top of them. Vs Paperclip: Ruflo = execution swarm inside tools; Paperclip = org/budget governance. Extreme ★ velocity — treat discovery heat as noise until the swarm loop proves out on your stack.","verdict":"MIT meta-harness when you need an execution swarm on existing coding CLIs — not a substitute for org governance or a simpler single-agent CLI.","addedAt":"2026-09-26","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","coding-agents","active","fast-stars"],"capabilities":["swarm","harness","agents"],"surfaces":["Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":76673,"starsPerWeek":1064,"access":"free","archived":false,"repo":{"createdAt":"2025-06-02","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/ruflo","html":"/gems/ruflo","markdown":"/gems/ruflo.md"}},{"slug":"cumora","name":"Cumora","tagline":"MIT team chat where AI agents are first-class teammates","category":"agents","url":"https://cumora.ai","github":"https://github.com/yetone/cumora","pricing":"open-source","hook":"Cumora is cross-platform team chat where AI agents sit on the roster as first-class teammates — DMs, groups, Kanban, calendar — via Cumora Cloud pods or BYOA on your machine.","whatsGood":"MIT (yetone/cumora). Roster/DMs/groups/Kanban/calendar for humans+agents. BYOA via `npx cumora agent computer` with Claude Code/Codex/etc so the server never sees provider keys; Cumora Cloud pods as the hosted path. Site at cumora.ai. ★3,889 from GitHub API 2026-09-27.","theCatch":"Vs Buzz: Nostr event-log workspace. Vs Multica: issue board. Vs OpenBot: per-bot computers. Vs OpenClaw: messaging employee. Hosted cloud path vs self-run BYOA — OSS repo is the chat+agent product; expect product heat around the cloud lane.","verdict":"MIT human+agent team chat when roster/Kanban/calendar beats another board or messaging bot — watch the Cloud vs BYOA split.","addedAt":"2026-09-27","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["self-host","local-first","permissive","active","young-repo"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":3996,"starsPerWeek":440,"access":"free","archived":false,"repo":{"createdAt":"2026-08-17","pushedAt":"2026-10-16","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/cumora","html":"/gems/cumora","markdown":"/gems/cumora.md"}},{"slug":"openshell","name":"OpenShell","tagline":"Apache-2.0 agents/sandbox policy runtime — deny-by-default kernel governance","category":"agents","url":"https://docs.nvidia.com/openshell","github":"https://github.com/NVIDIA/OpenShell","pricing":"open-source","hook":"OpenShell is NVIDIA’s open sandbox runtime for autonomous agents — declarative YAML policy from filesystem and process down to network and credential injection, without handing the agent the whole host.","whatsGood":"Apache-2.0 (NVIDIA/OpenShell). Agent-first sandboxes with Landlock/kernel-shaped isolation, hot-reloadable network policy, and provider-credential injection only at profile-authorized endpoints. Install via `install.sh` or PyPI `openshell`. Docs at docs.nvidia.com/openshell; compute drivers cover Docker, Podman, MicroVM, and Kubernetes. Ships agent skills for CLI, gateway debug, and policy generation.","theCatch":"Vs E2B / Daytona / CubeSandbox: those are hosted Firecracker sandboxes, freemium stateful env computers, or self-hosted RustVMM/KVM density — OpenShell’s catch is deny-by-default kernel governance + credential injection as policy, not a sandbox SaaS seat. Telemetry defaults on (disableable). ★10,805 in about 60 weeks (~180 a week), so treat the star count with caution; rank on substance, not the counter.","verdict":"Apache-2.0 policy runtime when agents need deny-by-default sandbox governance — not E2B/Daytona cloud seats; rank the YAML, not the ★.","addedAt":"2026-09-30","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","sandbox","active"],"capabilities":["sandbox","policy","agents"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":16647,"starsPerWeek":490,"access":"free","archived":false,"repo":{"createdAt":"2026-02-24","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/openshell","html":"/gems/openshell","markdown":"/gems/openshell.md"}},{"slug":"herdr","name":"Herdr","tagline":"Apache-2.0 terminal runtime for multi-agent herds — state, panes, socket API","category":"agents","url":"https://herdr.dev","github":"https://github.com/herdrdev/herdr","pricing":"open-source","hook":"Herdr is the Rust terminal runtime coding agents live on — workspaces, tiled panes, idle/busy/blocked detection, detach that keeps work alive, and a socket API agents use to herd each other.","whatsGood":"Apache-2.0 (herdrdev/herdr). Single binary in your existing terminal (Ghostty/Alacritty/Kitty/WezTerm/tmux). Detects Claude Code, Codex, Pi, OpenCode, Droid, Amp, and more. Agents spawn panes, prompt peers, and wait on real blocked state via CLI/socket. Multi-machine window, plugins marketplace, brew/curl install. Docs at herdr.dev/docs.","theCatch":"Vs Thurbox / Zellij / ECC: Thurbox is a lighter MIT tmux TUI orchestrator; Zellij is a general multiplexer; ECC is skills/memory/security on host agents — Herdr’s catch is agent-aware runtime + socket API, not another bare multiplexer and not an ecosystem pack. Agent state heuristics vary by harness; after machine restart, layout restores but original processes do not survive.","verdict":"Apache agent-terminal runtime when you herd many CLIs and need blocked/idle signals — not Thurbox-lite tmux, not Zellij alone.","addedAt":"2026-09-30","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["api","self-host","permissive","coding-agents","active","fast-stars"],"capabilities":["terminal","multiplexer","agents"],"surfaces":["API","Self-host"],"ecosystem":["Rust","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":44680,"starsPerWeek":1514,"access":"free","archived":false,"repo":{"createdAt":"2026-03-27","pushedAt":"2026-10-18","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/herdr","html":"/gems/herdr","markdown":"/gems/herdr.md"}},{"slug":"zeron","name":"Zeron","tagline":"MIT local control plane for coding agents — optional multi-device sync","category":"agents","url":"https://zeron.sh","github":"https://github.com/zeronsh/zeron","pricing":"open-source","hook":"Zeron is a local-first control plane for Claude Code, Codex, Cursor, Devin, Grok, Pi, and friends — drive sessions from a sidebar, keep an always-on box working, sync devices only when you opt in.","whatsGood":"MIT (zeronsh/zeron). Default local-only mode: no account, no network required. Per-device engine stores sessions on that machine; optional login for multi-device sync and remote drive. Linux curl installer + daemon; macOS desktop; Windows setup.exe. Follow/drive agents across devices when a VPS stays up.","theCatch":"Vs Paperclip / Herdr: Paperclip is org Board/budget governance; Herdr is the terminal multiplexer runtime — Zeron’s catch is cross-device session control UI/daemon, not budgets and not pane tiling. Synced devices trust each other with workspace file access (including gitignored files if enabled) — only sign in machines you trust. Sync does not migrate old local sessions automatically.","verdict":"MIT agent control plane when you want local-first multi-device drive of host agents — not Paperclip governance, not Herdr panes; treat sync as full workspace trust.","addedAt":"2026-09-30","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","local-first","permissive","coding-agents","active","young-repo"],"capabilities":["control-plane","sync","agents"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":3300,"starsPerWeek":252,"access":"free","archived":false,"repo":{"createdAt":"2026-07-20","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/zeron","html":"/gems/zeron","markdown":"/gems/zeron.md"}},{"slug":"nasiko","name":"Nasiko","tagline":"Apache-2.0 developer control plane — discover, route, and trace agent harnesses","category":"agents","url":"https://nasiko.ai","github":"https://github.com/Nasiko-Labs/nasiko","pricing":"open-source","hook":"Nasiko is a developer control plane for AI agent harnesses — discover what is already on the machine, deploy/route traffic, manage tools, and watch traces from one dashboard.","whatsGood":"Apache-2.0 (Nasiko-Labs/nasiko). Rust CLI (`nasiko agents discover` is read-only, no account, stays on-box). Dashboard for deploy, routing, tools, and traces. Hooks/plugins for OpenCode/Codex-style harnesses so operators keep one control surface while agents keep their native CLIs.","theCatch":"Vs Paperclip / Zeron / ECC: Paperclip is Board/budget company OS; Zeron is multi-device session drive; ECC is skills/memory/security pack — Nasiko’s catch is harness discovery + routing/trace control plane. Still early operator UX; confirm which harnesses are first-class before you bet a fleet on it.","verdict":"Apache harness control plane when you need discover/route/trace over many coding agents — not Paperclip budgets, not Zeron device sync.","addedAt":"2026-09-30","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["web-ui","self-host","permissive","license-unclear","active"],"capabilities":["control-plane","routing","agents"],"surfaces":["Self-host"],"ecosystem":["Rust","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":9469,"starsPerWeek":266,"access":"free","archived":false,"repo":{"createdAt":"2026-02-12","pushedAt":"2026-10-16","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/nasiko","html":"/gems/nasiko","markdown":"/gems/nasiko.md"}},{"slug":"monty","name":"Monty (Pydantic)","tagline":"MIT in-process Rust Python sandbox — OSS Monty only","category":"agents","url":"https://github.com/pydantic/monty","github":"https://github.com/pydantic/monty","pricing":"open-source","hook":"Monty is Pydantic’s open in-process Rust Python sandbox — run untrusted Python beside the agent without spinning a whole VM SaaS for every eval.","whatsGood":"MIT (pydantic/monty). OSS Monty only: `pydantic-monty` / `@pydantic/monty` / crates. Full Monty commercial is a separate product and is not required for the OSS bar. In-process Rust-backed Python sandbox aimed at agent tool runs and eval harnesses.","theCatch":"★8499 in about 41 weeks (~207 a week): judge it on the sandbox shape, not the star count. Vs E2B / AIO Sandbox / CubeSandbox: those are hosted Firecracker seats, broader sandbox/browser/MCP runtimes, or self-hosted RustVMM/KVM density — Monty’s catch is in-process Rust Python sandbox from Pydantic, not a sandbox SaaS seat. Do not confuse OSS Monty packages with Full Monty commercial; pin the MIT crates/npm/PyPI names.","verdict":"MIT in-process Python sandbox when agents need Pydantic Monty OSS — not E2B/AIO/CubeSandbox cloud or KVM seats; ignore the ★ spike.","addedAt":"2026-10-01","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","evals","sandbox","active"],"capabilities":["sandbox","python","rust"],"surfaces":["Self-host","Hosted"],"ecosystem":["Rust","Python","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":8719,"starsPerWeek":49,"access":"free","archived":false,"repo":{"createdAt":"2023-05-28","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/monty","html":"/gems/monty","markdown":"/gems/monty.md"}},{"slug":"claude-code-router","name":"Claude Code Router","tagline":"MIT local control plane — route coding agents across models","category":"agents","url":"https://github.com/musistudio/claude-code-router","github":"https://github.com/musistudio/claude-code-router","pricing":"open-source","hook":"Claude Code Router is a MIT local control plane — install `@musistudio/claude-code-router` and route coding-agent traffic across models with a CLI and web management UI.","whatsGood":"MIT (musistudio/claude-code-router). npm `@musistudio/claude-code-router` (HTTP 200) — not the unrelated bare `claude-code-router` package on npm. Local gateway with CLI + web UI so one agent seat can fuse/route providers without rewriting harness config each time.","theCatch":"★37492 in about 83 weeks (~451 a week): judge it on the router/gateway loop, not the star count. Vs weave-router / harnessrouter: Weave is ELv2 on-box embedder/prompt routing; HarnessRouter CE is an Apache unified API over coding harnesses — CCR’s catch is a local Claude-Code-oriented model router gateway + UI. Pin the scoped `@musistudio/` package name; bare npm name is a different project.","verdict":"MIT local agent model router when you want CCR’s gateway/UI — not Weave ELv2, not HarnessRouter; install `@musistudio/claude-code-router` only.","addedAt":"2026-10-01","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["cli","self-host","permissive","coding-agents","active"],"capabilities":["routing","gateway","cli"],"surfaces":["CLI","Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":37746,"starsPerWeek":439,"access":"free","archived":false,"repo":{"createdAt":"2025-02-25","pushedAt":"2026-10-10","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/claude-code-router","html":"/gems/claude-code-router","markdown":"/gems/claude-code-router.md"}},{"slug":"langflow","name":"Langflow","tagline":"MIT visual agent/workflow builder — every flow as API + MCP tool","category":"agents","url":"https://www.langflow.org","github":"https://github.com/langflow-ai/langflow","pricing":"open-source","hook":"Langflow is a MIT visual builder for AI agents and workflows — author in the UI, customize components in Python, and expose every flow as an API and MCP server tool.","whatsGood":"MIT (langflow-ai/langflow). PyPI `langflow` (HTTP 200). Visual authoring plus source-access Python components, playground, and batteries for major LLMs/vector DBs. Built-in API and MCP servers turn workflows into tools other agents can call.","theCatch":"★155k: a large, mature project, but stars still are not a quality signal. Vs Dify / n8n / mcp-agent: Dify is OSS+cloud LLM studio; n8n is fair-code general automation; mcp-agent is an agent framework — Langflow’s catch is the agent/RAG visual builder with first-class MCP tool export. Self-host ops and dependency weight are real; pin OSS deploy docs before promising zero-ops.","verdict":"MIT visual agent builder when flows must ship as API/MCP tools — not Dify’s studio lane alone and not n8n general automation.","addedAt":"2026-10-01","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","api","self-host","permissive","workflows","active","fast-stars"],"capabilities":["agents","workflows","mcp"],"surfaces":["MCP","API","Self-host"],"ecosystem":["Python","Multi-platform"],"licenseModel":"Open-Core","githubStars":155817,"starsPerWeek":808,"access":"free","archived":false,"repo":{"createdAt":"2023-02-08","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/langflow","html":"/gems/langflow","markdown":"/gems/langflow.md"}},{"slug":"oh-my-claudecode","name":"oh-my-claudecode","tagline":"MIT teams multi-agent orchestration for Claude Code — npm oh-my-claude-sisyphus","category":"agents","url":"https://github.com/Yeachan-Heo/oh-my-claudecode","github":"https://github.com/Yeachan-Heo/oh-my-claudecode","pricing":"open-source","hook":"oh-my-claudecode (OMC) is MIT teams-first multi-agent orchestration for Claude Code — install `oh-my-claude-sisyphus` and run orchestrated agent teams without learning every Claude Code knob.","whatsGood":"MIT (Yeachan-Heo/oh-my-claudecode). npm `oh-my-claude-sisyphus` (HTTP 200) — the package name is the install string. Multi-agent orchestration workflows aimed at Claude Code users who want teams without a custom harness. Sibling oh-my-codex exists for Codex CLI.","theCatch":"★39496 in about 38 weeks (~1040 a week): judge it on the orchestration substance, not the star count. Vs oh-my-hermes / openclaw / claude-code: oh-my-hermes is Hermes-agent niche memory/workflows; OpenClaw is multi-channel personal/coding agent — OMC’s catch is Claude Code multi-agent team orchestration via the sisyphus npm package. Package name ≠ repo name; pin `oh-my-claude-sisyphus` and verify which Claude Code versions the hooks target.","verdict":"MIT Claude Code team orchestration when OMC’s loops fit — not oh-my-hermes, not OpenClaw; install oh-my-claude-sisyphus; treat the star spike with caution.","addedAt":"2026-10-01","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","permissive","coding-agents","workflows","active","fast-stars"],"capabilities":["orchestration","claude-code","multi-agent"],"surfaces":["Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":40053,"starsPerWeek":989,"access":"free","archived":false,"repo":{"createdAt":"2026-01-09","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/oh-my-claudecode","html":"/gems/oh-my-claudecode","markdown":"/gems/oh-my-claudecode.md"}},{"slug":"kagent","name":"kagent","tagline":"Apache Kubernetes-native framework for building AI agents","category":"agents","url":"https://kagent.dev","github":"https://github.com/kagent-dev/kagent","pricing":"open-source","hook":"kagent is an Apache Kubernetes-native framework for building, deploying, and managing AI agents where your workloads already run — agents as first-class K8s citizens.","whatsGood":"Apache-2.0 (kagent-dev/kagent). Cloud-native agent framework with UI for agent ops, active releases, Discord, and OpenSSF best-practices badge. Homepage kagent.dev. Aimed at platform teams who want agents on the same orchestration plane as the rest of the stack.","theCatch":"Vs google-adk / mastra / langgraph / agentgateway: ADK/Mastra/LangGraph are code-first agent frameworks; Agentgateway is an MCP/A2A proxy — kagent’s catch is Kubernetes-native agent build/deploy/manage. You need a real cluster and K8s fluency; not the right first tool for a laptop coding-agent seat.","verdict":"Apache K8s-native agents when fleet deploy lives in the cluster — not ADK/Mastra app frameworks and not Agentgateway proxy alone.","addedAt":"2026-10-02","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["sdk","self-host","permissive","kubernetes","active"],"capabilities":["kubernetes","agents","cloud-native"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":4050,"starsPerWeek":45,"access":"free","archived":false,"repo":{"createdAt":"2025-01-21","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/kagent","html":"/gems/kagent","markdown":"/gems/kagent.md"}},{"slug":"gh-aw","name":"GitHub Agentic Workflows","tagline":"MIT gh extension — Markdown agent jobs compiled to Actions with safe outputs","category":"agents","url":"https://github.github.com/gh-aw/","github":"https://github.com/github/gh-aw","pricing":"open-source","hook":"GitHub Agentic Workflows (`gh-aw`) lets you write repo automation as Markdown with YAML frontmatter — issue triage, PR review, CI-failure investigation, docs upkeep — and compiles it into a normal GitHub Actions workflow where the agent runs read-only and sandboxed.","whatsGood":"MIT (github/gh-aw, latest v0.89.21). `gh extension install github/gh-aw`; `gh aw compile` validates the source and emits a `.lock.yml`. Engines: GitHub Copilot, Claude Code, OpenAI Codex, Google Gemini, Pi. Writes go through validated `safe-outputs` jobs with scoped permissions instead of giving the agent a write token. Ships agent-readable docs (llms.txt, llms-full.txt). ★~5.3k.","theCatch":"Maturity: pre-1.0 and the README itself says use with caution at your own risk; releases `>=0.83.3 <0.85.4` were retired after a security advisory — pin a current version. Cost is real: every run spends Actions minutes plus your engine’s tokens/seat. Permissions, tools, and network access are configurable, so a careless frontmatter undoes the sandbox. Vs claude-code-action: that runs Claude Code on PR/issue events; gh-aw is a multi-engine workflow compiler with a write-buffer safety model.","verdict":"MIT agentic GitHub Actions when you want triage/review/CI-investigation bots with read-only agents and buffered writes — pin versions and budget minutes plus tokens.","addedAt":"2026-10-05","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Orchestration","tags":["self-host","permissive","sandbox","workflows","active"],"capabilities":["ci"],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":5419,"starsPerWeek":87,"access":"free","archived":false,"repo":{"createdAt":"2025-08-12","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/gh-aw","html":"/gems/gh-aw","markdown":"/gems/gh-aw.md"}},{"slug":"docker-agent","name":"Docker Agent","tagline":"Apache-2.0 YAML-defined agents with MCP toolsets — runs as `docker agent`","category":"agents","url":"https://docker.github.io/docker-agent/","github":"https://github.com/docker/docker-agent","pricing":"open-source","hook":"Docker Agent (formerly cagent) lets you declare an agent — model, instructions, MCP toolsets, sub-agents — in one YAML file and run it with `docker agent run agent.yaml`, so agents version and share like a Compose file.","whatsGood":"Apache-2.0 (docker/docker-agent; the old docker/cagent URL redirects). Release v1.149.0 (2026-10-07). Pre-installed as a CLI plugin in Docker Desktop 4.63+, or `brew install docker-agent`. Built-in tools plus any MCP server (local, remote, or Docker-based), multi-agent orchestration, TUI, MCP mode, and RAG. Provider-agnostic: OpenAI, Anthropic, Gemini, Bedrock, Mistral, xAI, or local models via Docker Model Runner. ★~3.8k.","theCatch":"Agents are YAML config, not code in an SDK — great when your tools already ship as Docker images or MCP servers, limiting when you need custom control flow in Python/TypeScript. Local models need Docker Model Runner; otherwise bring your own provider API key and pay per token. Vs crewai (Python role/crew framework) and strands-agents (AWS model-driven SDK): Docker Agent is the declarative, container-native lane.","verdict":"Apache-2.0 declarative agent runner when your tools already live in Docker/MCP — YAML over SDK code; BYO key or Docker Model Runner.","addedAt":"2026-10-08","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["mcp","self-host","permissive","containers","active"],"capabilities":["yaml","mcp","docker"],"surfaces":["MCP","Self-host"],"ecosystem":["Go","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":4562,"starsPerWeek":77,"access":"free","archived":false,"repo":{"createdAt":"2025-09-01","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/docker-agent","html":"/gems/docker-agent","markdown":"/gems/docker-agent.md"}},{"slug":"eve","name":"eve","tagline":"Apache-2.0 filesystem-first framework for durable AI agents — from Vercel","category":"agents","url":"https://eve.dev","github":"https://github.com/vercel/eve","pricing":"open-source","hook":"eve makes the folder the agent: `instructions.md`, `tools/`, `skills/`, `channels/` (HTTP, Slack, Discord) and `schedules/` sit in conventional places, so an agent project is easy to read, diff, and extend — `npx eve@latest init my-agent` scaffolds one with a terminal UI.","whatsGood":"Apache-2.0 (vercel/eve), npm `eve@0.74.0`. Durable execution, subagents, sandbox, memory, evals, and schedules are documented in the package itself (`node_modules/eve/docs`), so coding agents can read the docs locally. Real Vercel backing and daily commits. Web Chat with Sign in with Vercel; `npx eve deploy` when you want hosting. ★~5.5k.","theCatch":"Model strings default to Vercel AI Gateway IDs; direct OpenAI and Anthropic helpers are built in, but any other provider means installing an AI SDK provider package and editing `agent.ts`. Public beta under Vercel beta terms — expect API churn. ★ is climbing far faster than a repo created in June 2026 with 1,000+ open issues and pull requests would suggest — judge it on substance (Vercel backing, shipping cadence), not the badge. Vs mastra (TypeScript agent framework, provider-neutral) and voltagent (TS framework + observability console): eve is the filesystem-convention, Vercel-native lane.","verdict":"Apache-2.0 filesystem-first agent framework when you want a Vercel-native, readable agent project — Gateway by default, beta APIs, star heat ahead of maturity.","addedAt":"2026-10-08","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["tui","api","sdk","self-host","permissive","skills","workflows","active","young-repo"],"capabilities":["typescript","agents","durable"],"surfaces":["API","Self-host"],"ecosystem":["Node.js"],"licenseModel":"Apache-2.0","githubStars":5611,"starsPerWeek":313,"access":"free","archived":false,"repo":{"createdAt":"2026-06-16","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/eve","html":"/gems/eve","markdown":"/gems/eve.md"}},{"slug":"scion","name":"Scion","tagline":"Apache-2.0 platform for running coding agents as teams","category":"agents","url":"https://googlecloudplatform.github.io/scion/","github":"https://github.com/GoogleCloudPlatform/scion","pricing":"open-source","hook":"Scion runs Claude Code, Codex, OpenCode or Antigravity as a team of agents, each in its own container with its own configuration and credentials, and lets them delegate tasks and message each other while you follow along from a CLI or a web UI.","whatsGood":"Apache-2.0 (GoogleCloudPlatform/scion), very active. Each agent can get a shared workspace or its own git worktree or clone, so parallel agents don't trample each other's files. scion start, attach, message, logs and resume manage agents running in tmux sessions. It starts on your laptop and can move to a hosted Hub on Kubernetes or Cloud Run, and an A2A bridge plus MCP configuration connect it to other agent systems. Installs with a community Homebrew tap. ★~1.9k.","theCatch":"It needs a container runtime: Podman, Docker or Apple Container (Kubernetes or Cloud Run for hosted runs). There are no GitHub releases yet, only preview tags such as v0.3.0-preview.3, and the README warns that the project is evolving rapidly and APIs and configuration may change. It is from a Google Cloud repo but, in its own words, not an officially supported Google product. Every agent still runs on your own model keys and their token costs.","verdict":"The most complete open setup for trying agent teams today — pin a preview tag and expect your config to change.","addedAt":"2026-10-17","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["cli","web-ui","self-host","permissive","coding-agents","containers","active"],"capabilities":["multi-agent","containers","worktrees"],"surfaces":["CLI","Self-host"],"ecosystem":["Go","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":1976,"starsPerWeek":62,"access":"free","archived":false,"repo":{"createdAt":"2026-03-10","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/scion","html":"/gems/scion","markdown":"/gems/scion.md"}},{"slug":"e2b","name":"E2B","tagline":"Firecracker sandboxes for untrusted agent code","category":"agents","url":"https://e2b.dev","github":null,"pricing":"freemium","hook":"E2B gives agents a safe place to run untrusted code.","whatsGood":"When your agent writes code, where does it run? E2B provides Firecracker-based sandboxes—lightweight VMs that spin up in milliseconds, execute untrusted code, and tear down cleanly. The infrastructure lets agents be dangerous without risking your systems. SDKs for Python and TypeScript.","theCatch":"Hosted by default. You are sending code to E2B infrastructure. Self-hosting is possible but requires more work. Another external dependency in your agent stack.","verdict":"Safe sandboxes for agent-generated code. Hosted convenience.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["free-tier","sandbox"],"capabilities":["sandbox","code"],"surfaces":["Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":null,"starsPerWeek":null,"access":"free-tier","archived":false,"repo":null,"links":{"self":"/api/gems/e2b","html":"/gems/e2b","markdown":"/gems/e2b.md"}},{"slug":"litellm","name":"LiteLLM","tagline":"MIT OpenAI-compatible gateway to 100+ LLMs; Enterprise SSO paid","category":"agents","url":"https://www.litellm.ai","github":"https://github.com/BerriAI/litellm","pricing":"freemium","hook":"LiteLLM is the self-host OpenAI-format proxy when every agent SDK speaking a different provider API is the mess.","whatsGood":"MIT core SDK + proxy. One OpenAI-compatible API across 100+ providers. Virtual keys, spend tracking, budgets, fallbacks, load balancing. Docker/Helm self-host; no telemetry required when you run it yourself.","theCatch":"Vs OpenRouter/provider SDKs: self-host gateway you operate — not a hosted model marketplace. Enterprise/ tree and SSO/audit/RBAC are commercial; core proxy stays MIT. You still pay model providers.","verdict":"MIT LLM gateway. Self-host free; Enterprise SSO/governance is paid.","addedAt":"2026-09-04","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["api","sdk","self-host","free-tier","license-unclear","auth","active"],"capabilities":["llm","proxy","gateway"],"surfaces":["API","SDK","Self-host","Hosted"],"ecosystem":["Python","Multi-platform"],"licenseModel":"Freemium","githubStars":61525,"starsPerWeek":365,"access":"free-tier","archived":false,"repo":{"createdAt":"2023-07-27","pushedAt":"2026-10-19","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/litellm","html":"/gems/litellm","markdown":"/gems/litellm.md"}},{"slug":"windmill","name":"Windmill","tagline":"AGPLv3 scripts→workflows/UIs/agents; CE free self-host unlimited runs","category":"agents","url":"https://www.windmill.dev","github":"https://github.com/windmill-labs/windmill","pricing":"freemium","hook":"Windmill is scripts-as-jobs with auto UIs and agent steps when Temporal is the heavy cluster and n8n is the fair-code canvas.","whatsGood":"AGPLv3 core. Turn TS/Python/Go/Bash/SQL into webhooks, schedules, DAG flows, and apps. CE self-host: unlimited executions, 3 workspaces, 50 users. Fast Rust job engine; MCP/agent tooling in flows. Cloud also has a free OSS tier.","theCatch":"Vs Temporal/Hatchet/n8n: script+UI internal platform, not Temporal durable SDK alone or n8n canvas. CE caps (4 groups, 10 SSO users, no durable audit logs, limited Git sync); Enterprise from ~$120/mo + seats/compute. Binary CE includes proprietary bits; from-source without enterprise flag is AGPL.","verdict":"AGPLv3 workflow/agent platform. CE self-host free with caps; EE for audit/SAML/scale.","addedAt":"2026-09-04","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","free-tier","license-unclear","workflows","active"],"capabilities":["workflows","scripts","agents"],"surfaces":["Self-host","Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":18219,"starsPerWeek":78,"access":"free-tier","archived":false,"repo":{"createdAt":"2022-05-05","pushedAt":"2026-10-19","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/windmill","html":"/gems/windmill","markdown":"/gems/windmill.md"}},{"slug":"daytona","name":"Daytona","tagline":"Freemium agent sandboxes — stateful envs, SDK/API/CLI","category":"agents","url":"https://www.daytona.io","github":"https://github.com/daytona/clients","pricing":"freemium","hook":"Daytona gives AI agents isolated, stateful sandboxes — filesystem, Git, shell, and code execution via SDK/API/CLI.","whatsGood":"Hosted sandboxes optimized for agent code execution and long-running state (snapshots, previews, computer use). Official clients at daytona/clients (Python/TS/Go/…). Integrates with common agent frameworks. Faster path than rolling your own Firecracker fleet.","theCatch":"Vs E2B (Firecracker sandboxes): Daytona pitches fuller stateful dev/env computers; both are hosted-by-default infra. daytonaio/daytona public core stopped updates June 2026 (private codebase) — treat product as freemium SaaS; Nightona is the community AGPL fork if you need self-host OSS. Clients repo ≠ full platform source.","verdict":"Freemium agent sandbox cloud. Compare E2B; self-host via Nightona not upstream.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["cli","api","sdk","free-tier","license-unclear","sandbox","git","active","young-repo"],"capabilities":["sandbox","agents","sdk"],"surfaces":["CLI","API","SDK","Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":15,"starsPerWeek":1,"access":"free-tier","archived":false,"repo":{"createdAt":"2026-06-23","pushedAt":"2026-10-18","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/daytona","html":"/gems/daytona","markdown":"/gems/daytona.md"}},{"slug":"dify","name":"Dify","tagline":"Open-source LLM app/agent studio + cloud freemium (LangGenius)","category":"agents","url":"https://dify.ai","github":"https://github.com/langgenius/dify","pricing":"freemium","hook":"Dify is the collaborative LLM app and agent studio — workflows, RAG, tools, and model glue in one workspace you can self-host or take to cloud.","whatsGood":"Modified Apache-2.0 (langgenius/dify) with additional commercial conditions. Visual workflows, RAG pipelines, agent/tooling, and multi-model support on one studio surface. Docker self-host path; docs at docs.dify.ai. Cloud at cloud.dify.ai when you do not want to run the stack. Move from prototype to production without rewriting the product shell.","theCatch":"dual OSS/cloud: self-host is free under the OSS license; Dify Cloud and edition upgrades are the paid product path — and the LICENSE adds conditions if you offer Dify as multi-tenant SaaS. Vs Mastra / LangGraph (SDK-only orchestration): Dify is a full product studio UI, not a code-first agent framework you embed. Model providers still meter either way.","verdict":"OSS+cloud LLM studio. Full product workspace vs SDK-only; watch SaaS license clauses and Cloud bills.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Orchestration","subcategory":"Orchestration","tags":["self-host","free-tier","license-unclear","rag","workflows","active","fast-stars"],"capabilities":["workflows","rag","studio"],"surfaces":["Self-host","Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":158842,"starsPerWeek":864,"access":"free-tier","archived":false,"repo":{"createdAt":"2023-04-12","pushedAt":"2026-10-19","license":"NOASSERTION","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/dify","html":"/gems/dify","markdown":"/gems/dify.md"}}],"meta":{"total":59,"count":59,"offset":0,"limit":null,"sort":"curated","sorts":["curated","catalog","new","stars","trending","name"],"filters":{"subcategory":"orchestration"},"next":null,"categories":["agents","nextjs","python","shopify","databases","devops","design","observability","rust","payments","comms","notifications","libraries"],"tags":["mcp","cli","tui","api","sdk","desktop-app","web-ui","editor-plugin","self-host","local-first","free-tier","paid-only","paid-led","permissive","copyleft","source-available","license-unclear","coding-agents","local-llm","rag","memory","evals","browser-automation","scraping","sandbox","skills","voice","security","auth","payments","realtime","postgres","vector","kubernetes","containers","observability","testing","workflows","graphics-3d","git","active","stale","archived","young-repo","fast-stars"]}}