egobot is a personal AI agent built by Citro Labs. It automates tasks and manages workflows across your devices using a local-first architecture — your data, login sessions, and workflow history stay on your own machine. No account required, no cloud sessions, no vendor lock-in.

egobot connects to your existing tools through a lightweight command-line interface. It uses semantic snapshots of application state rather than raw DOM or screen pixels, which makes interactions faster and dramatically reduces token usage. Successful task patterns are saved as reusable skills that improve with each run.

Any agent that can run a shell command can drive egobot — Claude Code, OpenAI Codex, Cursor, Gemini CLI, Kiro, Continue, Hermes Agent, OpenClaw, and any custom agent you build. Drop-in skills install in seconds, and there is no SDK or proprietary integration layer.

egobot runs tasks in parallel, each in an isolated context, without interfering with your active work. Its self-evolving skill system means repeat tasks skip exploration — up to 10x faster than the first run. And it uses compact accessibility-tree snapshots instead of raw HTML, which can reduce token usage by up to 90%.

Yes. egobot is designed with a local-first architecture. Browsing data, credentials, and task outputs remain on your device. The agent operates in an isolated workspace, and cross-context actions require explicit permission. Credentials stored in the auth vault never appear in agent context or command output.

egobot is currently available on macOS, with Windows and Linux support on the roadmap. The agent interface is platform-independent, so your skills and workflows will carry over when new platforms become available.