{"data":[{"slug":"stagehand","name":"Stagehand","tagline":"AI-native browser automation that mixes code and natural language","category":"agents","url":"https://stagehand.dev","github":"https://github.com/browserbase/stagehand","pricing":"open-source","hook":"Stagehand lets you write Playwright scripts, then swap in natural language when the selectors break.","whatsGood":"Stagehand solves the hardest problem in browser automation: what happens when the UI changes. Built on Playwright, it adds AI primitives—act, extract, observe—that interpret pages like a human would. You can mix deterministic code for the stable parts and natural-language instructions for the parts that shift.","theCatch":"AI calls add latency and cost. The natural-language parts are only as reliable as the underlying model. Debugging AI-driven steps is harder than debugging selectors.","verdict":"The right abstraction for browser automation that needs to survive UI changes.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["self-host","permissive","browser-automation","active"],"capabilities":["browser","playwright"],"surfaces":["Self-host"],"ecosystem":["Node.js","Browser"],"licenseModel":"OSS/MIT","githubStars":25729,"starsPerWeek":192,"access":"free","archived":false,"repo":{"createdAt":"2024-03-24","pushedAt":"2026-10-18","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/stagehand","html":"/gems/stagehand","markdown":"/gems/stagehand.md"}},{"slug":"browser-use","name":"Browser Use","tagline":"Open-source framework for LLM-driven browser agents","category":"agents","url":"https://browser-use.com","github":"https://github.com/browser-use/browser-use","pricing":"open-source","hook":"Browser Use turns any LLM into a browser agent that navigates, clicks, and extracts like a human.","whatsGood":"Browser Use is the open-source default for teams building AI agents that need to use the web. Point it at a URL, describe what you want, and the LLM handles navigation, form filling, and data extraction. The framework abstracts browser control into a clean API that works with multiple model providers.","theCatch":"You need browser infrastructure—Browser Use is a framework, not a managed service. Token consumption scales with page complexity. Anti-bot measures will block you without stealth tooling.","verdict":"The open-source foundation for browser-based AI agents.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["sdk","self-host","permissive","browser-automation","active","fast-stars"],"capabilities":["browser","llm"],"surfaces":["Self-host"],"ecosystem":["Python"],"licenseModel":"OSS/MIT","githubStars":118448,"starsPerWeek":1154,"access":"free","archived":false,"repo":{"createdAt":"2024-10-31","pushedAt":"2026-10-17","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/browser-use","html":"/gems/browser-use","markdown":"/gems/browser-use.md"}},{"slug":"playwright-mcp","name":"Playwright MCP","tagline":"Official Microsoft MCP server for Playwright browser automation","category":"agents","url":"https://github.com/microsoft/playwright-mcp","github":"https://github.com/microsoft/playwright-mcp","pricing":"open-source","hook":"Playwright MCP gives Claude Desktop and every MCP client real browser control.","whatsGood":"Microsoft’s official MCP server that exposes Playwright to any MCP-compatible tool. Claude Desktop, Cursor, Windsurf, Cline—any agent that speaks MCP can now navigate, click, and extract from the web. Accessibility snapshots provide structured DOM perception instead of pixel-based selectors.","theCatch":"It is a control interface, not a complete solution. You still need browser infrastructure (local Chromium or Browserbase). No built-in session management, authentication, or scaling.","verdict":"The standard way to give MCP clients browser access.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["mcp","desktop-app","self-host","permissive","browser-automation","active"],"capabilities":["mcp","browser"],"surfaces":["MCP","Self-host"],"ecosystem":["Node.js","Browser"],"licenseModel":"Apache-2.0","githubStars":38351,"starsPerWeek":465,"access":"free","archived":false,"repo":{"createdAt":"2025-03-21","pushedAt":"2026-10-16","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/playwright-mcp","html":"/gems/playwright-mcp","markdown":"/gems/playwright-mcp.md"}},{"slug":"chrome-devtools-mcp","name":"Chrome DevTools MCP","tagline":"Official Chrome DevTools as an MCP server for coding agents","category":"agents","url":"https://github.com/ChromeDevTools/chrome-devtools-mcp","github":"https://github.com/ChromeDevTools/chrome-devtools-mcp","pricing":"open-source","hook":"Chrome DevTools MCP gives Cursor, Claude, and Copilot a real Chrome browser with DevTools attached.","whatsGood":"Official Chrome DevTools team MCP server. Agents get performance traces, network inspection, console with source-mapped stacks, screenshots, and Puppeteer-backed automation that waits for results. Works with Antigravity, Claude, Cursor, Copilot. Slim mode for basic browsing; full mode for debugging. This is the browser surface built for coding agents, not a generic scraper.","theCatch":"chrome-devtools-mcp@1.8.0 engines: \"^20.19.0 || ^22.12.0 || >=23\". 20.16 and 20.17 fail; 20.19 LTS is in engines. “Node ≥ 22.12” and “Node 20 fails” are both too blunt — README “LTS” is marketing; engines is the catch. A CLI is provided for use without MCP; official CLI docs call it experimental. Complementary to Playwright MCP —DevTools-grade traces, Lighthouse, heap snapshots, and --autoConnect, not a generic scraper or a Playwright replacement. Officially supports Google Chrome and Chrome for Testing only. Exposes browser content to the MCP client—do not point it at sensitive sessions. Usage statistics on by default (opt out with --no-usage-statistics). Needs a current Chrome.","verdict":"The MCP browser for agents that need real DevTools, not just clicks.","addedAt":"2026-09-03","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["mcp","self-host","permissive","coding-agents","browser-automation","active","fast-stars"],"capabilities":["mcp","devtools","browser"],"surfaces":["MCP","Self-host"],"ecosystem":["Node.js","Browser"],"licenseModel":"Apache-2.0","githubStars":53813,"starsPerWeek":933,"access":"free","archived":false,"repo":{"createdAt":"2025-09-11","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/chrome-devtools-mcp","html":"/gems/chrome-devtools-mcp","markdown":"/gems/chrome-devtools-mcp.md"}},{"slug":"agent-browser","name":"agent-browser","tagline":"Vercel Labs Rust CLI with compact a11y refs for agents","category":"agents","url":"https://agent-browser.dev","github":"https://github.com/vercel-labs/agent-browser","pricing":"open-source","hook":"agent-browser snapshots a compact accessibility tree and lets the agent click `@e1` instead of a CSS selector.","whatsGood":"Vercel Labs Rust CLI (Apache-2.0). Native daemon talks CDP to Chrome. `agent-browser snapshot -i` returns an a11y tree with refs (`@e1`, `@e2`); click/fill use those refs. Distinct from Browser Use (Python LLM browser framework), Playwright MCP (MCP control plane), and Chrome DevTools MCP (DevTools traces/Lighthouse). CLI-first for any agent that can run a shell—Claude Code, Cursor, Codex, Gemini. Native binaries for macOS, Linux, Windows.","theCatch":"Default idle timeout is about one hour: the headless daemon shuts down and discards transient tabs/state unless you passed `--restore` (or another restore key). Headed, Safari/iOS WebDriver, and user-attached browsers are exempt; cloud-provider browsers are not. Tune with `--idle-timeout` or `AGENT_BROWSER_IDLE_TIMEOUT_MS`; `0` disables. This is CDP Chrome, not stealth—anti-bot pages will still block you.","verdict":"The CLI browser for agents that want refs, not a full DOM dump. Restore or lose the session.","addedAt":"2026-09-03","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["cli","self-host","permissive","browser-automation","active","fast-stars"],"capabilities":["browser","a11y","cli"],"surfaces":["CLI","Self-host"],"ecosystem":["Rust"],"licenseModel":"Apache-2.0","githubStars":44210,"starsPerWeek":1099,"access":"free","archived":false,"repo":{"createdAt":"2026-01-11","pushedAt":"2026-10-18","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/agent-browser","html":"/gems/agent-browser","markdown":"/gems/agent-browser.md"}},{"slug":"steel","name":"Steel","tagline":"Apache-2.0 browser API for AI agents — self-host or cloud","category":"agents","url":"https://steel.dev","github":"https://github.com/steel-dev/steel-browser","pricing":"open-source","hook":"Steel is the open-source browser API for agents: sessions, CDP, stealth, and scrape helpers without owning Chromium ops.","whatsGood":"Apache-2.0 (steel-dev/steel-browser). Docker one-liner for API+UI; Puppeteer/Playwright/Selenium over CDP; session state, proxies, extensions, markdown/screenshot/PDF helpers. Cloud at steel.dev when you do not want to run Chrome fleets. Built for agent automation, not generic SEO scrapers only.","theCatch":"Vs Browserbase (managed fleet), Stagehand (AI Playwright layer), Browser Use (LLM browser framework): Steel is the browser sandbox/API — bring your own agent logic. Self-host means Chrome resource cost; cloud is freemium/paid convenience. Anti-bot arms race still applies.","verdict":"Apache-2.0 agent browser API. Self-host Steel; pay Browserbase-shaped cloud only if you must.","addedAt":"2026-09-14","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["api","self-host","permissive","browser-automation","scraping","active"],"capabilities":["browser","api","agents"],"surfaces":["API","Self-host"],"ecosystem":["Node.js","Browser","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":7835,"starsPerWeek":76,"access":"free","archived":false,"repo":{"createdAt":"2024-11-01","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/steel","html":"/gems/steel","markdown":"/gems/steel.md"}},{"slug":"scrapling","name":"Scrapling","tagline":"BSD-3 adaptive scrape/crawl framework + MCP","category":"agents","url":"https://scrapling.readthedocs.io","github":"https://github.com/D4Vinci/Scrapling","pricing":"open-source","hook":"Scrapling is an adaptive Python scrape/crawl framework — from one request to full-scale crawls — with MCP for agents.","whatsGood":"BSD-3-Clause (D4Vinci/Scrapling). Handles fetching, parsing, and crawl scale with an adaptive stance toward modern sites. Docs at scrapling.readthedocs.io; Docker Hub image; MCP surface so agents can call scrape tools without inventing another HTTP client. pip-installable Python stack.","theCatch":"Vs Firecrawl-style hosted scrape APIs: Scrapling is a library/framework you run (plus MCP), not a metered scrape SaaS with a managed anti-bot fleet. You own proxies, blockers, and ops; MCP does not erase site ToS or legal risk. Not a general browser-agent product like Browser Use/Steel.","verdict":"BSD-3 adaptive scrape/crawl + MCP. Self-run framework vs Firecrawl-shaped APIs.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["mcp","sdk","self-host","permissive","browser-automation","scraping","active","fast-stars"],"capabilities":["scrape","crawl","mcp"],"surfaces":["MCP","Self-host"],"ecosystem":["Python"],"licenseModel":"BSD-3-Clause","githubStars":88529,"starsPerWeek":841,"access":"free","archived":false,"repo":{"createdAt":"2024-10-13","pushedAt":"2026-10-18","license":"BSD-3-Clause","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/scrapling","html":"/gems/scrapling","markdown":"/gems/scrapling.md"}},{"slug":"agent-reach","name":"Agent Reach","tagline":"MIT eyes for agents — public-web CLI, zero API fees","category":"agents","url":"https://github.com/Panniantong/Agent-Reach","github":"https://github.com/Panniantong/Agent-Reach","pricing":"open-source","hook":"Agent Reach gives AI agents eyes on the public web — read and search Twitter, Reddit, YouTube, GitHub, and more via one CLI with zero API fees.","whatsGood":"MIT (Panniantong/Agent-Reach). One CLI aimed at agent install/configure/health-check flows so you are not hand-wiring every site SDK. Public-web reach without buying search/scrape API keys for day-one use. MCP-friendly install path documented in-repo.","theCatch":"Vs Tavily / Exa (paid search APIs): Agent Reach is CLI eyes on public surfaces without a search SaaS bill — not a ranked enterprise search product with SLAs. Site ToS, auth/cookie setup, and breakage are yours; some platforms need user-controlled sessions. Not a replacement for licensed data APIs.","verdict":"MIT agent web CLI. Zero API-fee eyes; not Tavily/Exa-grade paid search.","addedAt":"2026-09-15","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["cli","api","self-host","permissive","browser-automation","active","fast-stars"],"capabilities":["cli","search","web"],"surfaces":["CLI","API","Self-host"],"ecosystem":["Python","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":99181,"starsPerWeek":2922,"access":"free","archived":false,"repo":{"createdAt":"2026-02-24","pushedAt":"2026-10-15","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/agent-reach","html":"/gems/agent-reach","markdown":"/gems/agent-reach.md"}},{"slug":"browserskill","name":"BrowserSkill","tagline":"Attach coding agents to a real logged-in browser without interrupting the session","category":"agents","url":"https://github.com/Tencent/BrowserSkill","github":"https://github.com/Tencent/BrowserSkill","pricing":"open-source","hook":"BrowserSkill attaches a coding agent to a real, already-logged-in browser session so it can work with the browser you are actually using.","whatsGood":"MIT (Tencent/BrowserSkill). The useful wedge is session continuity: an agent can operate in a live browser context instead of requiring a separate automation instance and a second login. That makes authenticated, stateful browser work more practical for coding agents.","theCatch":"A live logged-in browser is also a live blast radius: the agent inherits cookies, tabs, permissions, and whatever account is open. Vs Stagehand, Browser Use, and Playwright MCP (separate browser instances or contexts): BrowserSkill trades isolation for continuity. Treat the attached session as the trust boundary, not as a sandbox.","verdict":"MIT live-session browser bridge. The no-second-login wedge is real; inherited cookies and permissions are the Catch.","addedAt":"2026-09-18","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["self-host","permissive","coding-agents","browser-automation","active","young-repo","fast-stars"],"capabilities":["browser","agents","session"],"surfaces":["Self-host"],"ecosystem":["Browser","Multi-platform"],"licenseModel":"OSS/MIT","githubStars":9172,"starsPerWeek":537,"access":"free","archived":false,"repo":{"createdAt":"2026-06-22","pushedAt":"2026-10-15","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/browserskill","html":"/gems/browserskill","markdown":"/gems/browserskill.md"}},{"slug":"typesafe-computer-use","name":"typesafe-computer-use","tagline":"MIT macOS computer-use loop — OCR + TypeSafe Jev decisions, ~$0.0002/step","category":"agents","url":"https://github.com/awlevin/typesafe-computer-use","github":"https://github.com/awlevin/typesafe-computer-use","pricing":"open-source","hook":"typesafe-computer-use drives a Mac toward a plain-English goal without shipping screenshots to a frontier model — OCR the screen, classify the next action with TypeSafe Jev, click.","whatsGood":"MIT (awlevin/typesafe-computer-use). Deterministic screen read + calibrated Choice classifier instead of full multimodal computer-use. Author-measured ~$0.0002 per decision and ~1.5s end-to-end steps vs multi-second frontier screenshot loops. Python 3.12+ / uv / macOS 14+. Honest README: rebuild frontier free reasoning as deterministic state.","theCatch":"The decision brain is closed TypeSafe Jev (paid input tokens, early-access/waitlist) — the loop is OSS, the classifier is not. macOS only. Vs Open Interpreter / browser-use / SemIf: this is GUI clicker + paid System One decisions, not a coding REPL, browser agent, or local logit SemIf gate. A separate writing model is still needed when a text field needs free text.","verdict":"Cheap macOS computer-use around TypeSafe Jev. OSS loop; paid/waitlisted decision API is the real dependency.","addedAt":"2026-09-19","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Browser Automation","tags":["self-host","permissive","browser-automation","active","young-repo"],"capabilities":[],"surfaces":["Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"OSS/MIT","githubStars":1243,"starsPerWeek":259,"access":"free","archived":false,"repo":{"createdAt":"2026-09-16","pushedAt":"2026-09-29","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/typesafe-computer-use","html":"/gems/typesafe-computer-use","markdown":"/gems/typesafe-computer-use.md"}},{"slug":"cua","name":"Cua","tagline":"Open computer-use drivers, sandboxes, MCP, and CUA-S1 decision models","category":"agents","url":"https://cua.ai","github":"https://github.com/trycua/cua","pricing":"open-source","hook":"Cua gives agents a computer-use control plane: OS and desktop drivers, sandboxes, MCP, and CUA-S1 decision models for acting beyond a browser tab.","whatsGood":"MIT (trycua/cua). The stack covers computer-use drivers across OS/desktop surfaces, local sandboxes, MCP integration, and the CUA-S1 model line when an agent needs a decision loop rather than a text-only tool call. It is a useful bridge from browser automation to the wider desktop.","theCatch":"Vs BrowserSkill, Stagehand, and Browser Use: Cua is the broader computer-use driver/sandbox/model stack, not only browser automation or a browser-session attach bridge. Optional `omni` extras pull AGPL Ultralytics; cloud fleets are paid. Audit extras and treat a fleet as infrastructure spend, not as part of the MIT core.","verdict":"MIT computer-use stack with CUA-S1. Reach for desktop/OS control and sandboxes; audit the AGPL omni extra and paid cloud fleet boundary.","addedAt":"2026-09-20","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["mcp","desktop-app","self-host","copyleft","browser-automation","sandbox","active"],"capabilities":["computer-use","browser","mcp"],"surfaces":["MCP","Self-host"],"ecosystem":["Python","Multi-platform"],"licenseModel":"AGPL-3.0","githubStars":30167,"starsPerWeek":337,"access":"free","archived":false,"repo":{"createdAt":"2025-01-31","pushedAt":"2026-10-19","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/cua","html":"/gems/cua","markdown":"/gems/cua.md"}},{"slug":"jev-ultrafast","name":"Jev Ultrafast","tagline":"Browser Use × TypeSafe Jev — indexed action-space web agent","category":"agents","url":"https://github.com/browser-use/jev-ultrafast","github":"https://github.com/browser-use/jev-ultrafast","pricing":"freemium","hook":"Jev Ultrafast is Browser Use’s fast browser agent: TypeSafe Jev picks operation + element from a live control table; a small LLM only writes text when the action is TYPE_TEXT.","whatsGood":"MIT (browser-use/jev-ultrafast). One TypeSafe request per decision cycle with speculative target heads, no screenshots in the default loop, and a local inspector at :8766. Demo claim: Zürich→London on Google Flights in ~7.1s including typing and waits. Library API (`Agent(url, goal)`) plus uv-run CLI; Chrome connects through Browser Harness.","theCatch":"Vs Browser Use and typesafe-computer-use: this is a Jev-driven indexed web agent, not the general Browser Use framework and not a macOS desktop OCR loop. Needs TYPESAFE_API_KEY plus a text-model key (OpenRouter in the example); Browser Use Cloud ultrafast is waitlist. Demos stop short of booking — verify outcomes yourself.","verdict":"MIT Browser Use × Jev speed agent when you want structured click/type loops without screenshot multimodal; budget TypeSafe + text-model keys.","addedAt":"2026-09-23","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Browser Automation","tags":["free-tier","permissive","browser-automation","active","young-repo","fast-stars"],"capabilities":["browser"],"surfaces":["Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":23237,"starsPerWeek":4843,"access":"free","archived":false,"repo":{"createdAt":"2026-09-16","pushedAt":"2026-09-30","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/jev-ultrafast","html":"/gems/jev-ultrafast","markdown":"/gems/jev-ultrafast.md"}},{"slug":"moli","name":"Moli","tagline":"Apache/MIT Rust headless browser for AI agents — structure first","category":"agents","url":"https://github.com/lexmount/moli","github":"https://github.com/lexmount/moli","pricing":"open-source","hook":"Moli is a production-oriented headless browser built for AI agents: on-demand layout/rendering, low footprint, and control via CLI, CDP, WebDriver Classic, or WebDriver BiDi.","whatsGood":"Dual Apache-2.0 or MIT at your option (lexmount/moli). Rust runtime aimed at fetch/extract/search/automate with skills under the repo; Linux/macOS/Windows. Benchmarks claim roughly ~15% of Chrome CPU and ~13% peak memory on their headless card — useful when agents need a browser without Chrome’s weight.","theCatch":"Vs agent-browser and Playwright: Moli is its own lightweight browser runtime, not a CDP wrapper around Chrome and not a general test framework. Default install is curl|sh / irm|iex of release installer scripts — verify checksums. No GUI/GPU compositor; not pixel-parity Chrome; Canvas/WebGL/media are intentional gaps.","verdict":"Dual-license agent browser when you want a slim Rust runtime with CDP/WebDriver surfaces; keep Chrome-based tools if you need full web fidelity.","addedAt":"2026-09-24","featured":false,"domain":"AI Agents","subSpecialty":"Agents","subcategory":"Browser Automation","tags":["cli","self-host","permissive","browser-automation","graphics-3d","active","young-repo","fast-stars"],"capabilities":["cli","browser"],"surfaces":["CLI","Self-host"],"ecosystem":["Multi-platform"],"licenseModel":"Apache-2.0","githubStars":19521,"starsPerWeek":1936,"access":"free","archived":false,"repo":{"createdAt":"2026-08-10","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/moli","html":"/gems/moli","markdown":"/gems/moli.md"}},{"slug":"workflow-use","name":"Workflow Use","tagline":"AGPL RPA 2.0 — record/run durable browser workflows","category":"agents","url":"https://github.com/browser-use/workflow-use","github":"https://github.com/browser-use/workflow-use","pricing":"open-source","hook":"Workflow Use is Browser Use’s RPA 2.0 layer: record or generate deterministic browser workflows with variables, then rerun them — falling back to Browser Use when a step fails.","whatsGood":"AGPL-3.0 (browser-use/workflow-use). Recorder / generation mode → reusable semantic workflows stored locally; CLI to generate, list, and run stored workflows. Sibling of Browser Use — the point is durable workflow/RPA, not another open-ended browser agent.","theCatch":"Vs Browser Use: pitch this as the workflow/RPA layer on that lineage, not a second browser agent. AGPL-3.0. README says very early development / not production; optional Browser Use Cloud browsers are convenience, not required for the OSS path.","verdict":"AGPL browser workflow/RPA when you need record→rerun determinism on Browser Use rails and can accept early-dev plus copyleft.","addedAt":"2026-09-25","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["self-host","copyleft","browser-automation","workflows","active"],"capabilities":["browser","workflows","rpa"],"surfaces":["Self-host"],"ecosystem":["Python","Browser"],"licenseModel":"AGPL-3.0","githubStars":4226,"starsPerWeek":56,"access":"free","archived":false,"repo":{"createdAt":"2025-05-06","pushedAt":"2026-10-16","license":"AGPL-3.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/workflow-use","html":"/gems/workflow-use","markdown":"/gems/workflow-use.md"}},{"slug":"invisible-playwright-mcp","name":"Invisible Playwright MCP","tagline":"MIT stealth Firefox Playwright MCP — not plain Chromium MCP","category":"agents","url":"https://github.com/feder-cr/invisible_playwright_mcp","github":"https://github.com/feder-cr/invisible_playwright_mcp","pricing":"open-source","hook":"Invisible Playwright MCP is a Playwright MCP server on anti-detect stealth Firefox so agents browse with less bot friction than plain Chromium MCP.","whatsGood":"MIT now (feder-cr/invisible_playwright_mcp; pre-2026-09-02 dists were AGPL). Free via `uvx` and Claude/Codex/Gemini plugins. MCP-first stealth Firefox layer when official Playwright MCP’s Chromium path gets blocked.","theCatch":"Vs official Playwright MCP: plain Chromium control — this is stealth Firefox. Vs Browser Use / Steel / Cua: those are agent frameworks or browser APIs; this is the MCP-first stealth layer. Anti-detect purpose → site ToS / abuse risk even though the license is clean.","verdict":"MIT stealth Firefox MCP when plain Playwright MCP dies on anti-bot — only with explicit ToS/abuse judgment, not as default browser MCP.","addedAt":"2026-09-26","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["mcp","self-host","copyleft","browser-automation","active","young-repo","fast-stars"],"capabilities":["mcp","stealth","browser"],"surfaces":["MCP","Self-host"],"ecosystem":["Python","Browser"],"licenseModel":"AGPL-3.0","githubStars":2773,"starsPerWeek":943,"access":"free","archived":false,"repo":{"createdAt":"2026-09-29","pushedAt":"2026-10-17","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/invisible-playwright-mcp","html":"/gems/invisible-playwright-mcp","markdown":"/gems/invisible-playwright-mcp.md"}},{"slug":"ui-tars-desktop","name":"UI-TARS Desktop","tagline":"Apache-2.0 vision/GUI-agent desktop stack (Agent TARS)","category":"agents","url":"https://agent-tars.com","github":"https://github.com/bytedance/UI-TARS-desktop","pricing":"open-source","hook":"UI-TARS Desktop is ByteDance’s Apache-2.0 multimodal GUI-agent stack: vision-first desktop/browser operators via Releases, brew, or Agent TARS CLI.","whatsGood":"Apache-2.0 (bytedance/UI-TARS-desktop). Free desktop app / brew / Agent TARS CLI. Vision/GUI-agent first — screens and clicks, not only DOM tools. Ships a real desktop GUI-agent product surface at agent-tars.com.","theCatch":"Vs Browser Use: LLM-DOM browser agents; UI-TARS is vision/GUI-agent first. Vs Cua: Cua is computer-use drivers/sandboxes/models; UI-TARS ships the desktop GUI-agent product. Needs a VLM (HF Apache-2.0 weights or VolcEngine/Doubao cloud); Remote Operator free trial ended 2025-08-20.","verdict":"Apache-2.0 GUI-agent desktop when you want vision operators — budget a VLM path; do not expect the ended Remote Operator trial.","addedAt":"2026-09-26","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["cli","desktop-app","self-host","permissive","browser-automation","active"],"capabilities":["gui","vision","desktop"],"surfaces":["CLI","Self-host"],"ecosystem":["Node.js","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":39319,"starsPerWeek":431,"access":"free","archived":false,"repo":{"createdAt":"2025-01-19","pushedAt":"2026-10-09","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/ui-tars-desktop","html":"/gems/ui-tars-desktop","markdown":"/gems/ui-tars-desktop.md"}},{"slug":"agent-desktop","name":"agent-desktop","tagline":"Apache-2.0 desktop computer-use CLI — OS a11y refs, not pixels","category":"agents","url":"https://github.com/lahfir/agent-desktop","github":"https://github.com/lahfir/agent-desktop","pricing":"open-source","hook":"agent-desktop gives agents real desktop computer use through OS accessibility trees — stable refs and retry-safe actions instead of screenshot guessing.","whatsGood":"Apache-2.0 (lahfir/agent-desktop). Rust CLI + C-ABI cdylib; progressive skeleton snapshots (claimed ~78–96% token cuts on dense UIs); ref-based click/type; headless-by-default a11y actions; CDP launch path so Playwright/agent-browser can drive Chromium apps while native chrome stays on a11y. npm `agent-desktop` / npx; ClawHub + skills.sh listings.","theCatch":"Production a11y drive is macOS-first — Windows/Linux a11y and input still Planned in the support matrix (cdylib ships wider). Vs agent-browser (Vercel Labs browser CDP/a11y refs): agent-browser is the web/browser CLI; agent-desktop is native desktop apps via OS accessibility. Needs Accessibility (+ Screen Recording/Automation for some features); do not hand it a logged-in machine without a sandbox story.","verdict":"Apache desktop computer-use when agents must drive native macOS apps by a11y refs — not agent-browser’s web CDP lane; Win/Linux still catching up.","addedAt":"2026-09-30","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["cli","desktop-app","self-host","permissive","browser-automation","active"],"capabilities":["desktop","a11y","computer-use"],"surfaces":["CLI","Self-host"],"ecosystem":["Rust","Multi-platform"],"licenseModel":"Apache-2.0","githubStars":1806,"starsPerWeek":52,"access":"free","archived":false,"repo":{"createdAt":"2026-02-19","pushedAt":"2026-10-19","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/agent-desktop","html":"/gems/agent-desktop","markdown":"/gems/agent-desktop.md"}},{"slug":"nanobrowser","name":"Nanobrowser","tagline":"Apache-2.0 Chrome extension that runs a multi-agent web automation loop with your own LLM keys","category":"agents","url":"https://nanobrowser.ai","github":"https://github.com/nanobrowser/nanobrowser","pricing":"open-source","hook":"Nanobrowser puts a Planner and a Navigator agent in a Chrome side panel: you type a task, they click and type through real pages in your own browser, using whatever model you point them at.","whatsGood":"Apache-2.0 (nanobrowser/nanobrowser), release v0.2.0 (2026-10-02). Free on the Chrome Web Store, or load the release zip unpacked for newer builds. Bring your own key for OpenAI, Anthropic, Gemini, DeepSeek, Grok, Azure OpenAI, OpenRouter, Ollama or any OpenAI-compatible endpoint, and pick a different model per agent. Follow-up questions and task history in the side panel. ★~14k.","theCatch":"Anonymous analytics (task metrics and visited domain names) are on by default — turn them off in settings. Chrome and Edge only. It drives your logged-in browser, so a hostile page can try to steer the agent with prompt injection while your sessions are live; use a separate profile for anything sensitive. You pay per token on your own key, and multi-step tasks burn more than you expect. Vs browser-use (Python library you script) and playwright-mcp (browser tools for an agent you already run): Nanobrowser is a point-and-ask extension, not something you build on.","verdict":"The free, bring-your-own-key Operator alternative for one-off web chores — switch off analytics and keep it in a separate browser profile.","addedAt":"2026-10-10","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["self-host","permissive","browser-automation","active"],"capabilities":["browser","extension","byok"],"surfaces":["Self-host"],"ecosystem":["Browser"],"licenseModel":"Apache-2.0","githubStars":14119,"starsPerWeek":150,"access":"free","archived":false,"repo":{"createdAt":"2024-12-31","pushedAt":"2026-10-15","license":"Apache-2.0","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/nanobrowser","html":"/gems/nanobrowser","markdown":"/gems/nanobrowser.md"}},{"slug":"mcp-chrome","name":"Chrome MCP Server","tagline":"MIT Chrome extension MCP server for your own browser","category":"agents","url":"https://github.com/hangwin/mcp-chrome","github":"https://github.com/hangwin/mcp-chrome","pricing":"open-source","hook":"mcp-chrome is a Chrome extension plus a small bridge (npm install -g mcp-chrome-bridge) that lets any MCP client control the Chrome you already use — with your logins, cookies and open tabs.","whatsGood":"MIT (hangwin/mcp-chrome), TypeScript, release v1.0.0 (2025-12-29). 20+ tools: screenshots, clicks and typing, network monitoring, bookmarks, history, and cross-tab context. Fully local, with Streamable HTTP transport and a built-in vector index for semantic search across open tabs. ★~12.5k.","theCatch":"Because it uses your real browser session, an agent can act as you on every site you are logged in to — point it at a separate Chrome profile for anything sensitive. The README still calls it early-stage, and the last commit was January 2026, so momentum has slowed. The extension is installed from GitHub releases, not the Chrome Web Store. Vs chrome-devtools-mcp (official from the Chrome DevTools team): that one gives DevTools-grade traces, Lighthouse and heap snapshots; mcp-chrome is about reusing your logged-in profile for everyday browsing tasks.","verdict":"The quickest way to let an agent work in your logged-in Chrome — use a separate profile and keep expectations early-stage.","addedAt":"2026-10-14","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["mcp","self-host","permissive","browser-automation","stale"],"capabilities":["mcp","browser","extension"],"surfaces":["MCP","Self-host"],"ecosystem":["Node.js","Browser"],"licenseModel":"OSS/MIT","githubStars":12497,"starsPerWeek":176,"access":"free","archived":false,"repo":{"createdAt":"2025-06-09","pushedAt":"2026-01-06","license":"MIT","archived":false,"checkedAt":"2026-10-19"},"links":{"self":"/api/gems/mcp-chrome","html":"/gems/mcp-chrome","markdown":"/gems/mcp-chrome.md"}},{"slug":"browserbase","name":"Browserbase","tagline":"Managed browser infrastructure for AI agents at scale","category":"agents","url":"https://browserbase.com","github":null,"pricing":"freemium","hook":"Browserbase runs the browsers so your agents can focus on the task.","whatsGood":"Browserbase handles the infrastructure AI agents need to use the web: isolated browser sessions, session recordings for debugging, browser profiles for persistent logins, and concurrency that scales. Connect with Playwright, Puppeteer, Stagehand, or Browser Use without managing your own Chromium fleet.","theCatch":"Another service to pay for and depend on. Complex pricing at scale—budget carefully for high-volume workloads. You are outsourcing control over the browser layer.","verdict":"The browser infrastructure that lets you focus on the agent, not the fleet.","addedAt":"2026-08-19","featured":false,"domain":"AI Agents","subSpecialty":"Browser Automation","subcategory":"Browser Automation","tags":["free-tier","paid-led","browser-automation"],"capabilities":["browser","infra"],"surfaces":["Hosted"],"ecosystem":["Multi-platform"],"licenseModel":"Freemium","githubStars":null,"starsPerWeek":null,"access":"paid-led","archived":false,"repo":null,"links":{"self":"/api/gems/browserbase","html":"/gems/browserbase","markdown":"/gems/browserbase.md"}}],"meta":{"total":20,"count":20,"offset":0,"limit":null,"sort":"curated","sorts":["curated","catalog","new","stars","trending","name"],"filters":{"subcategory":"browser-automation"},"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"]}}