Runbook AI

Spending too much time on repetitive web tasks? Runbook AI is a local-first AI agent that lives in your browser sidepanel. Give it a goal in plain language and watch it navigate, click, fill forms, and read pages in real time — using the same browser you're already signed into. What it can do: • Browse and automate: multi-step tasks across tabs and pages — searches, form fills, data extraction, following links, working with shadow DOM and iframes. • Schedule recurring runs: "check HN top 5 every morning", "send me a stock-price report twice a day". • Monitor pages for changes: cheap DOM-diff polling fires the agent only when content actually changes (Gmail inbox, PR status, dashboards, queues). • Native CLI delegation: hand off self-contained sub-tasks to Claude Code or Gemini CLI on your host without leaving the chat. • MCP bridge: invoke the agent from external coding tools and agents. Privacy by design: • Runs entirely in your browser — no Runbook-hosted servers in the loop. • Talks only to the LLM provider you configure (OpenAI, Anthropic, Gemini, local Ollama, and other OpenAI-compatible endpoints). • No eval() or remote-loaded code — enforced by the Chrome extension sandbox. • Your task results are stored locally within the extension, and will be deleted once you remove the extension. Discover and share runbooks at runbookai.net. Join the Discord community (https://discord.gg/SDtXkAKK2B) to share feedback and follow development. You can also invoke Runbook AI from Claude Code, Gemini CLI, and other coding agents via runbook-ai-mcp (https://github.com/runbook-ai/runbook-ai-mcp), an MCP bridge to the extension.