Best agent memory tools (2026)

1. 1.Mem0Memory infrastructure for AI agents that actually persists

   Free / OSS · Freemium · 67.4k GitHub stars

   VerdictThe memory layer most agent teams should use instead of building.

   CatchExternal dependency for a core capability. Async processing adds latency. Platform optimizations are not in the open-source SDK. You are trusting Mem0 with potentially sensitive user data.

2. 2.LangfuseOpen-source LLM observability you can self-host

   Free / OSS · Open-Core · 36k GitHub stars

   VerdictThe LLM observability platform for teams who want to own their data.

   CatchLicense: the core is MIT, but anything under the ee/ directories is under a separate enterprise license — check which features you rely on before calling your deployment fully MIT. Self-hosting means self-operating—you are running Postgres, ClickHouse, Redis, and the app. Observability-first means evals are layered on, not central.

3. 3.LettaEpisodic memory for agents that think across sessions

   Free / OSS · Apache-2.0 · 25.2k GitHub stars

   VerdictThe memory architecture for agents that need to think long-term.

   CatchMore complex to reason about than simple vector retrieval. The paging mechanism adds latency and orchestration overhead. Documentation assumes understanding of the original MemGPT paper.

4. 4.PromptfooMIT LLM evals & red-teaming CLI; Community free, 10k probes/mo

   Free / OSS · Freemium · 26.1k GitHub stars

   VerdictMIT eval/red-team CLI. Free core; probe cap + Enterprise sales wall.

   CatchVs DeepEval/Braintrust: YAML/CLI-first, not pytest-native Python or Braintrust SaaS. Community caps red-team probes (~10k/mo); Enterprise/On-Prem are contact-sales (SSO, dashboards, dedicated runner). LLM API costs are yours. Acquisition governance is a multi-year bet.

5. 5.DeepEvalApache-2.0 pytest-style LLM/agent evals; Confident AI is the paid cloud

   Free / OSS · Apache-2.0 · 18.9k GitHub stars

   VerdictApache-2.0 LLM unit tests. Free locally; Confident AI starts free then $200.

   CatchOSS is the runner — collaboration/dashboards/monitoring are Confident AI (Free: 2 seats/1 project/5 runs/wk; Starter $200/mo; Team $2k/mo). LLM-as-judge burns provider tokens. Vs Ragas: broader agent/chat coverage, not RAG-only.

6. 6.GraphitiApache-2.0 temporal knowledge graphs for agent memory; Zep's OSS engine

   Free / OSS · Apache-2.0 · 32k GitHub stars

   VerdictApache-2.0 temporal memory graphs. Free self-host; Zep Cloud is the paid path.

   CatchFramework only — you run a graph DB + ops. Zep Community Edition deprecated; full Zep SaaS is separate (Free credits then Flex ~$104/mo billed annually). Not a drop-in chat history store. Schema/ops cost vs Mem0 Docker.

7. 7.agentsviewMIT local-first agent session/cost archive — 20+ coding agents

   Free / OSS · OSS/MIT · 6.2k GitHub stars

   VerdictMIT local agent archive. Session/cost on disk; not Langfuse/Braintrust cloud.

   CatchVs Langfuse / Braintrust (cloud LLM observability): agentsview is a local session/cost archive for coding-agent transcripts, not a hosted tracing/eval platform. Optional anonymous telemetry ping; you still need the agent logs on disk. Not a replacement for production OpenTelemetry of your app’s LLM calls.

8. 8.EngramMIT persistent agent memory — MCP, CLI, TUI, SQLite

   Free / OSS · OSS/MIT · 7.4k GitHub stars

   VerdictMIT agent memory primitive. MCP/CLI/TUI on SQLite; not a full agent framework.

   CatchVs Mem0 / Letta / full agent frameworks: Engram is the memory primitive (store + retrieve + surfaces), not an orchestration framework or coding agent. “local or cloud” branding means optional hosted/sync gravity — day-one stays OSS you run. You still decide what gets written and when.

9. 9.oh-my-hermesMIT Hermes-agent niche tooling — memory & workflows

   Free / OSS · OSS/MIT · 3.4k GitHub stars

   VerdictMIT Hermes-agent tooling. Niche, not a multi-host skill pack.

   CatchHermes-specific — not a generic Claude Code / Codex skill pack. Young surface; expect Hermes Agent as the host and follow upstream Hermes breaking changes. Vs ECC (multi-host ecosystem) and security-audit-skill (Cloudflare audit skill): oh-my-hermes is niche Hermes tooling, not a cross-harness coordination layer or a security scanner.

10. 10.nanobotUltra-light self-hosted personal AI agent with WebUI, MCP, and memory

    Free / OSS · OSS/MIT · 49.2k GitHub stars

    VerdictMIT personal agent runtime. Self-host when you want MCP + memory + chat bridges in a small Python core.

    CatchYou operate the host, keys, and chat-app credentials; “ultra-light” still means a long-running service. Young surface with a large issue backlog — pin versions for production. Vs OpenHands / LibreChat (heavier coding or multi-model chat stacks) and OpenBot (governed coworker computers): nanobot is the small personal gateway, not a full browser-cowork platform or a repo-coding agent.

Also reviewed (15)

• Friday — Self-hosted cognitive memory MCP for coding agents
• SemIf — MIT local/WebGPU semantic conditionals — OpenJev, the former pitch name
• AX Check — Hosted free agent-experience checklist + recorded coding-agent sessions
• Supermemory — MIT memory/context engine with a local API
• OpenViking — AGPL self-evolving context database for memory, RAG, and skills
• OKF Agent Memory — Git-native dual-memory for agents — OKF v0.2 + embedded MCP
• ai-memory — Zero-LLM git-backed Markdown memory wiki with SQLite FTS
• OpenHuman — GPL-3 local-first agent orchestrator with memory trees and MCP
• book-to-skill — MIT converter from books/docs into cross-agent SKILL.md packs
• Hindsight — MIT retain→recall→reflect agent memory (+ MCP)
• Cognee — Apache-2.0 AI memory platform — local knowledge graph for agents
• iFixAi — Apache-2.0 agents eval/audit — pip install, premium ≠ paywall
• Obsidian Mind — MIT Obsidian vault as persistent agent memory
• Beads — MIT issue-tracker memory for coding agents, on Dolt
• Braintrust — Eval-first LLM platform that blocks bad prompts in CI