Zenii Documentation
Zenii is a local AI backend for developers who want one service behind their scripts, desktop UI, terminal tools, and MCP clients.
Start Here
- Installation & Usage — Get up and running in minutes
- API Reference — HTTP and WebSocket endpoints
- CLI Reference — Command-line usage and recipes
- Configuration — Runtime settings and defaults
- Polyglot Agent Runtime (PAR) — Run external Python/Node code as tools
- Architecture — System design and internals
- Deployment — Native, Docker, and server deployment
- Development — Build, test, and contribute
What Ships Today
- Daemon API — Local REST and WebSocket service at
http://localhost:18981(55+ routes) - Shipped interfaces — Desktop app, CLI, TUI, and MCP server
- Persistent memory — Shared local state across interfaces and automations
- Built-in tools — 16 base tools, with channels, scheduler, and workflows behind feature flags
- Polyglot Agent Runtime (PAR) — Run external Python (via
uv) and Node.js (vianpx) code as tools; self-healing on failure - Local-first deployment — Self-hosted, MIT licensed, zero telemetry in the product surface
Good Fit
- Developers building local automations, coding workflows, or agent tooling
- Teams that want one internal AI backend instead of separate per-tool integrations
- People who want a native desktop UI and a scriptable local API backed by the same service
Current Boundaries
- Zenii is not a hosted SaaS
- Zenii is not a drop-in OpenAI-compatible server today
- Mobile is planned, but not shipped in this repository
Interactive API Explorer
When running the Zenii daemon, visit localhost:18981/api-docs for the interactive Scalar API explorer with try-it-out functionality.