
Witness
MCP serverRecords your voice and a browser session, then hands agents a causal evidence package
Runs locally- 9 tools
- Open source
- Developer tools

Records your voice and a browser session, then hands agents a causal evidence package
Runs locallyIsolated browser profiles with their own fingerprint and proxy for each agent task
Not checked yet
Persistent memory for Claude Code, Codex and OpenCode, stored in SQLite on your machine
Runs locallyTyped website functions for agents, as an alternative to driving a browser
Browser actions for AI agents that are classified, approved and logged
Runs locallyShare VS Code's built-in browser with Claude Code, Copilot, Codex and other agents
Runs locally
Write and play load test scripts, build scenarios and read results in Loadster
Give an assistant a cloud browser to scrape, search, crawl and audit pages
31 web scraping, search and research tools that return clean Markdown or JSON
Runs locallyCapture webpage screenshots and PDFs from your AI assistant with a single tool
Not checked yetCloud Android phones for agents to see, tap, type and install apps over MCP

Desktop GUI automation through accessibility trees, with replayable tests
Runs locallyMCP connector that hands tasks to a MyClawn cloud computer with its own browser and email
Not checked yetA visible desktop browser, one workspace per project, that coding agents drive over MCP
Runs locallyLocal npm runtime for Kadmo agents: workflows and Chrome automation on your computer
Runs locallyLet an agent read Safari history, tabs and reading list, with writes switched off
Runs locally
MCP server that drives your real Chrome in per-agent background tabs on Windows
Runs locallyCapture browser flows into reviewable Java tests and diagnose failures from evidence
Runs locallyMac app so Claude and Cursor can read Reddit, X and LinkedIn through your Chrome login
Runs locallyLocal MCP server for SnapRender screenshots, with a free trial key
Runs locallyTurns websites into typed JSON tool calls that agents can search, read, log in and act on
Not checked yetPersistent Linux cloud desktops that an agent can see, click, type into and run tasks on
Not checked yetCloud Android phones an AI agent can see, tap and drive, or hand a task in plain words
Not checked yetMCP server that lets an assistant operate any VNC-reachable desktop by screenshot
Runs locallyBrowser automation over raw CDP in one Python file, with an MCP server for AI agents
Runs locallyCreate, launch and drive Nullprint antidetect browser profiles from an AI agent
Runs locallyBrowser MCP that records website routes and replays them without model calls
Runs locallyRun 56 web performance snippets in headless Chromium and read Core Web Vitals results
Runs locallySelf-hosted backend that turns a database into secured APIs and MCP tools
Not checked yetFetch bot-protected and JavaScript-heavy pages as Markdown or template-parsed JSON
Not checked yet