

Canopy
Quick Start
Last updated
1. Start the daemon
canopy daemon start
The daemon owns the MCP server (Streamable HTTP on port 7755 + stdio), the cron scheduler, the file watcher engine and the database.
2. Configure (first time)
canopy setup
Choose your AI CLI platforms (Claude Code, and others from the registry), model flags, temperature units and RAG settings.
3. Launch the TUI
canopy
Opens the full-screen Canopy Hub. Press Ctrl+N (or n) to create a
new agent:
- Interactive — a live PTY session with your chosen AI CLI.
- Background — cron-scheduled or file-watcher-triggered.
- Terminal — a raw shell session with per-session history.
Optionally select a seed identity to give the agent a persistent personality.
4. Wire up your AI CLI via MCP
canopy mcp
The MCP wizard syncs, adds and removes canopy’s MCP server entry across
all detected platforms, converting formats automatically (JSON ↔ TOML).
Add offers three input modes: Paste JSON (the snippet from the
server’s README, installed verbatim under its own name), Local command
(stdio) (command with its arguments plus optional KEY=VALUE env vars),
and Remote URL (http) (URL plus optional Name=Value HTTP headers).
Entries are previewed before writing — overwrites are marked
(replaces existing) — and one confirm installs them everywhere.
See the CLI reference for details.
After that, your AI agents can call all
84 MCP tools — schedule background agents, store
knowledge, coordinate with each other.
5. Verify
canopy doctor # full health diagnostics
canopy daemon status # is the daemon running?
canopy daemon logs # tail the daemon log
Where to go next
- Agents — the three agent kinds and seed identities.
- Intelligence & Sync — persistent knowledge.
- Graphs — multi-step automation.
- RAG Pipeline — index and search your documents.