What is Zeno
A personal agent that operates across the apps you already use, by composing the connectors you install.
Zeno is a personal agent that acts inside the apps you already work in — opens a pull request after you fix a Sentry error, triages your inbox, lists the issues blocking a sprint, comments on a PR with a code-review summary. The mission is action across systems, not chat.
Mental model
Zeno has four moving parts. Once you understand the layering, everything else in these docs falls into place.
- The agent. A reasoning loop that takes a request from a channel, decides which tools to call, and produces a reply. The agent has no shell, no filesystem of its own, and no general web access — every external capability flows through a connector.
- Connectors. Small MCP servers the operator installs through the dashboard. Each connector exposes typed tools (
mcp__github-app__merge_pull_request,mcp__sentry__list_issues, etc.) the agent composes. Connectors are the heart of the product — without them, Zeno is a talking statue. - Channels. The I/O boundary. A channel is how a request gets to Zeno and how the reply goes back out. Slack today.
- The backend. The reasoning engine — Claude Code via the OAuth-authenticated SDK. The backend talks to connectors through the agent.
Who it is for
- Single operator. One person per Zeno install. Multi-user (teams, allowlists, billing) is explicitly out of scope.
- Self-hosted. Zeno runs in Docker on your machine. There is no hosted SaaS.
- Experimental. Personal project, no SLA. Breaking changes are expected; the release line uses CalVer.
If that fits, install Zeno. Otherwise, the project's GitHub repo carries the code and the public roadmap.