Browser tools

• StagehandAI-native browser automation that mixes code and natural language

  Free / OSS · OSS/MIT · 25.8k GitHub stars

  VerdictThe right abstraction for browser automation that needs to survive UI changes.

  CatchAI 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.

• Browser UseOpen-source framework for LLM-driven browser agents

  Free / OSS · OSS/MIT · 119k GitHub stars

  VerdictThe open-source foundation for browser-based AI agents.

  CatchYou 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.

• Playwright MCPOfficial Microsoft MCP server for Playwright browser automation

  Free / OSS · Apache-2.0 · 38.4k GitHub stars

  VerdictThe standard way to give MCP clients browser access.

  CatchIt 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.

• Chrome DevTools MCPOfficial Chrome DevTools as an MCP server for coding agents

  Free / OSS · Apache-2.0 · 53.9k GitHub stars

  VerdictThe MCP browser for agents that need real DevTools, not just clicks.

  Catchchrome-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.

• agent-browserVercel Labs Rust CLI with compact a11y refs for agents

  Free / OSS · Apache-2.0 · 44.4k GitHub stars

  VerdictThe CLI browser for agents that want refs, not a full DOM dump. Restore or lose the session.

  CatchDefault 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.

• ColyseusMIT Node multiplayer rooms with state sync

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

  VerdictMIT multiplayer rooms. Self-host is the free path; Cloud is paid from dollar one.

  CatchA server framework, not a renderer. Colyseus Cloud has no free tier — it starts at 15 USD/mo. Capacity depends on your code, not a marketed CCU cap.

• PlaywrightApache-2.0 cross-browser library and CLI

  Free / OSS · Apache-2.0 · 98.2k GitHub stars

  VerdictApache-2.0 browser library. Free; MCP wrapper is a separate gem.

  CatchVs Playwright MCP gem: library/CLI lane; not the MCP wrapper. Heavier than unit tests. Browser downloads and CI images are real ops. Not Stagehand/Browserbase.

• HoppscotchMIT web-first open-source API client ecosystem

  Free / OSS · OSS/MIT · 80.7k GitHub stars

  VerdictMIT API client. Free core + self-host; pick it when the browser is the client, Bruno when Git is the source of truth.

  CatchVs Bruno (git-native *desktop* collections as files): Hoppscotch is web/self-host first, not “collections are just commits” by default. Cloud/workspace features and self-host ops are the real product surface — Enterprise exists for orgs. Not a mock/monitor platform.

• MQTT.jsMIT Node + browser MQTT client — needs a broker

  Free / OSS · OSS/MIT · 9.1k GitHub stars

  VerdictMIT MQTT client. Free library; bring a broker — browsers speak WebSockets.

  CatchClient library only — not a broker (that is Mosquitto/HiveMQ/etc.). Browser is MQTT-over-WebSockets only (no raw MQTT TCP from the page). You still provide a reachable broker, credentials, and topic design. Distinct from Mosquitto.

• SteelApache-2.0 browser API for AI agents — self-host or cloud

  Free / OSS · Apache-2.0 · 7.8k GitHub stars

  VerdictApache-2.0 agent browser API. Self-host Steel; pay Browserbase-shaped cloud only if you must.

  CatchVs 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.

• OpenBotSelf-hosted AI coworkers with per-bot computers and action governance

  Free / OSS · OSS/MIT · 6.8k GitHub stars

  VerdictMIT AG-UI coworker platform. Clone when you need audited agent computers on your infra; still alpha.

  CatchAlpha template, not a hosted product — expect rough edges and you operate Compose yourself. Single-user default is for laptop demos; turn on real sign-in before multi-user. Vs CopilotKit (agent UI/protocol layer) and nanobot (personal gateway): OpenBot is the governed coworker computer stack built on AG-UI, not the React UI kit alone.

• BrowserSkillAttach coding agents to a real logged-in browser without interrupting the session

  Free / OSS · OSS/MIT · 9.3k GitHub stars

  VerdictMIT live-session browser bridge. The no-second-login wedge is real; inherited cookies and permissions are the Catch.

  CatchA 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.

• CuaOpen computer-use drivers, sandboxes, MCP, and CUA-S1 decision models

  Free / OSS · AGPL-3.0 · 30.5k GitHub stars

  VerdictMIT computer-use stack with CUA-S1. Reach for desktop/OS control and sandboxes; audit the AGPL omni extra and paid cloud fleet boundary.

  CatchVs 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.

• Jev UltrafastBrowser Use × TypeSafe Jev — indexed action-space web agent

  Free / OSS · Freemium · 23.4k GitHub stars

  VerdictMIT Browser Use × Jev speed agent when you want structured click/type loops without screenshot multimodal; budget TypeSafe + text-model keys.

  CatchVs 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.

• OpenMuseMIT personal agent app — browser, terminal, files, lasting work

  Free / OSS · OSS/MIT · 5.4k GitHub stars

  VerdictMIT personal-agent starter when you want visible browser/files work on your box; treat live Google + Intelligence paths as optional, configured layers.

  CatchVs CopilotKit core and OpenClaw-class assistants: OpenMuse is the personal-agent application template, not the horizontal UI SDK alone and not a multi-tenant SaaS. Alpha — live Google accounts, CopilotKit Intelligence Rich Threads, and real models need their own keys/config; Node 24 + pnpm 11 pinned. Not multi-user auth.

• ZCodeApache coding workspace — desktop, web, and zcode Agent CLI

  Free / OSS · Apache-2.0 · 8.1k GitHub stars

  VerdictApache coding workspace when you want desktop+web+CLI from one OSS tree and can accept Z.ai product edges.

  CatchVs Claude Code, Pi, and OpenCode: ZCode is Z.ai’s full workspace stack, not Anthropic’s product CLI, Earendil’s harness, or a pure terminal OSS agent. Bootstrap is heavy (pinned Node/pnpm, desktop/runtime assets); some OAuth/product routes still point at Z.ai services. Chinese/English docs mix — English README.en.md is the clearer entry.

• AgentVerse OSApache personal cloud OS for you + coding agents on one Ubuntu box

  Free / OSS · Apache-2.0 · 981 GitHub stars

  VerdictApache personal agent cloud when you will run Ubuntu+Tailscale yourself; skip if you wanted a managed IDE or a simple CLI harness.

  CatchVs Coder, Umbrel, and Pi: this is a full personal cloud OS for agents, not a workspace control plane alone, a home NAS app store alone, or a coding harness. Alpha 0.2 on a single-person test box — no multi-user accounts yet; needs Ubuntu + ZFS/Incus/Tailscale ops weight. Bring your own Claude Code/Codex subscriptions inside workspaces.

• MoliApache/MIT Rust headless browser for AI agents — structure first

  Free / OSS · Apache-2.0 · 19.9k GitHub stars

  VerdictDual-license agent browser when you want a slim Rust runtime with CDP/WebDriver surfaces; keep Chrome-based tools if you need full web fidelity.

  CatchVs 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.

• AIO SandboxApache one container: browser+shell+files+VS Code+Jupyter+MCP

  Free / OSS · Apache-2.0 · 6.2k GitHub stars

  VerdictApache 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.

  CatchVs 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.

• Workflow UseAGPL RPA 2.0 — record/run durable browser workflows

  Free / OSS · AGPL-3.0 · 4.2k GitHub stars

  VerdictAGPL browser workflow/RPA when you need record→rerun determinism on Browser Use rails and can accept early-dev plus copyleft.

  CatchVs 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.

• BrowserOSAGPL agentic browser — open alternative to Atlas/Comet/Dia

  Free / OSS · AGPL-3.0 · 14k GitHub stars

  VerdictAGPL agentic browser when you want an Atlas/Comet-class shell you can build from source — not another browser-automation library.

  CatchVs Browser Use, agent-browser, BrowserSkill, and Chrome DevTools MCP: this is a forked agentic browser product, not a library/CLI that drives Chrome you already run. AGPL-3.0 — redistribution and SaaS wrapping need copyleft diligence. Chromium fork ops/update burden is yours.

• Invisible Playwright MCPMIT stealth Firefox Playwright MCP — not plain Chromium MCP

  Free / OSS · AGPL-3.0 · 1.3k GitHub stars

  VerdictMIT stealth Firefox MCP when plain Playwright MCP dies on anti-bot — only with explicit ToS/abuse judgment, not as default browser MCP.

  CatchVs 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.

• Qwen CodeApache-2.0 multi-surface coding agent (Qwen lineage)

  Free / OSS · Apache-2.0 · 28.6k GitHub stars

  VerdictApache-2.0 multi-surface coding agent when Qwen-lineage open models + Skills/Teams/MCP beat a closed product CLI.

  CatchVs Claude Code: paid product CLI. Vs OpenCode / ZCode / fx: other OSS coding agents — Qwen Code is the Qwen-lineage multi-surface agent with open models+framework, not a tiny Zig binary or Earendil harness fork.

• gstackMIT Garry Tan’s Claude Code role stack — plan, review, QA, ship

  Free / OSS · OSS/MIT · 137k GitHub stars

  VerdictMIT opinionated Claude Code software factory when a solo founder wants plan→review→QA→ship roles out of the box — heavy setup and real-session browser power; pin it and audit what it can log into.

  CatchBlast radius: browser skills drive your real logged-in sessions (Aside on macOS) or import cookies from Chrome/Brave/Arc into its bundled browser — treat `/setup-browser-cookies` and the sidebar agent like handing over your accounts. 135k stars in ~7 months — rank on the workflows, not the counter. Claude Code is the only “full” tier; other hosts are advisory, and outside reviews expect Codex/Claude CLIs signed in. Team mode auto-updates every session (throttled hourly) — pin if you need reproducible agents. Vs mattpocock-skills / agent-skills / superpowers: those are smaller composable packs; gstack is a whole opinionated org chart with a browser runtime.

• PlannotatorMIT/Apache local browser review for agent plans, docs and diffs — feedback goes straight back to the agent

  Free / OSS · Apache-2.0 · 9.5k GitHub stars

  VerdictThe best free way to actually read and mark up an agent plan before it codes — local, multi-harness, open source; mind the installer footprint and the release check.

  CatchEvery review surface checks GitHub for a newer release when it loads and there is no switch to turn that off; local code review also runs git ls-remote against origin unless you disable it. The full installer edits config for every agent it detects, so use --minimal or read what it touched. Integration depth varies per harness (Droid gets commands only, no plan interception yet). It adds a browser round-trip to every plan, which is great for big changes and friction for one-liners.

• OmnigentApache-2.0 meta-harness over Claude Code, Codex, Cursor, Pi and custom agents — sessions on any device, policies, sandboxes

  Free / OSS · Apache-2.0 · 10.9k GitHub stars

  VerdictMost complete open-source way to run several coding agents under one roof with approvals and sandboxes — powerful but young, heavy to install, and telemetry-on by default.

  CatchAnonymized usage telemetry is on by default; the opt-out lives in their docs, so flip it before you point it at work code. Heavy toolchain: Python 3.12+, uv, git, Node 22 with pnpm for the web UI, tmux for terminal wrappers, and bubblewrap is mandatory on Linux. First commit was June 2026, so APIs and harness support move fast. A meta-harness is a second layer to debug when an agent misbehaves. Vs Herdr (terminal runtime for agent herds): Omnigent is the bigger server-plus-web-UI play with policies and remote sandboxes.

• cmuxNative macOS terminal for parallel coding agents

  Free / OSS · GPL-3.0 · 28.3k GitHub stars

  VerdictGPL native macOS terminal when you run many coding agents in parallel and keep missing the one that is waiting — server pieces are BUSL, and it is Mac-only.

  CatchmacOS only. The license is split: the app is GPL, but the server parts (web/ and several workers/ directories) are Business Source License 1.1 and need a commercial license from Manaflow for production use, including self-hosting. Browser import can pull cookies and sessions from 20+ browsers into its panes — convenient, and a lot of trust to hand a terminal. Over 3,000 open issues and pull requests on a fast-moving repo. Vs ghostty (the emulator it builds on) and thurbox (tmux TUI orchestrator): cmux is the native GUI for agent attention, not a muxer or a harness.

• CrawleeApache-2.0 Node.js crawling and scraping library from Apify, HTTP or headless browser

  Free / OSS · Apache-2.0 · 26.2k GitHub stars

  VerdictThe default Node scraping framework when a one-off Playwright script has turned into a crawler — budget for proxies.

  CatchNeeds Node 22.13 or newer. The Python version is a separate project with its own feature set, so do not assume parity. Heavy production crawling usually needs paid proxies to avoid blocks — the library rotates them, it does not supply them. Vs apify-mcp (gives an agent access to ready-made Apify scrapers) and playwright (the browser automation underneath): Crawlee is the crawler framework you build your own scraper on.

• NanobrowserApache-2.0 Chrome extension that runs a multi-agent web automation loop with your own LLM keys

  Free / OSS · Apache-2.0 · 14.1k GitHub stars

  VerdictThe free, bring-your-own-key Operator alternative for one-off web chores — switch off analytics and keep it in a separate browser profile.

  CatchAnonymous 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.

• PGliteApache-2.0 Postgres compiled to WASM for browser, Node, Bun and Deno

  Free / OSS · Apache-2.0 · 16.2k GitHub stars

  VerdictThe easiest way to get real Postgres in tests, the browser or a local-first app — not your production server.

  CatchIt is single user and single connection, so it is not a replacement for a Postgres server behind a multi-user backend. Performance and memory are bounded by the WASM runtime, and only the extensions the project has ported can be loaded. It is still 0.x, so expect API changes between minor versions.

More with this tag (19)

• Nginx UI — AGPL-3.0 web UI for managing Nginx configs, certs and logs
• Backrest — GPL-3.0 web UI and scheduler for restic backups
• PrivateBin — Zlib zero-knowledge encrypted pastebin you self-host
• SeleniumBase — MIT Python browser automation and testing
• Dexie.js — Apache-2.0 friendly IndexedDB wrapper for the browser
• TypeORM — MIT TypeScript ORM with Active Record and Data Mapper
• drawDB — AGPL-3.0 browser ER diagram editor and SQL generator
• Orama — Apache-2.0 full-text, vector and hybrid search in JS
• Kubetail — Apache-2.0 real-time log dashboard for Kubernetes workloads
• libGDX — Apache-2.0 cross-platform Java game framework
• Artillery — MPL-2.0 load testing for HTTP, WebSockets and browsers
• JSON Crack — Apache-2.0 JSON visualizer that turns data into graphs
• ttyd — MIT share your terminal over the web
• Skyvern — AGPL-3.0 browser automation driven by an LLM plus vision, not selectors
• Lightpanda — AGPL-3.0 headless browser written in Zig for agents and scraping
• PeerJS — MIT browser P2P WebRTC with free cloud or self-host signaling
• Chrome MCP Server — MIT Chrome extension MCP server for your own browser
• Browserbase — Managed browser infrastructure for AI agents at scale
• Figma — Collaborative design tool in the browser