Daemon Engine (AI-First)

Agent-oriented runtime designed so AI agents can operate and maintain it themselves.

Quickstart (agent-friendly)

• Prereqs: Node.js 20+, npm.
• Install deps: npm install
• Build: npm run build
• Test: npm test
• Lint + types: npm run check
• Run (built output): node dist/main.js or node bin/daemon-engine.js --config path/to/daemon.yaml

Configuration

• Default search order: ./daemon.yaml, ./daemon.json, ~/.config/daemon-engine/daemon.yaml, ~/.config/daemon-engine/daemon.json.
• Key fields:
  • workspace: path to workspace dir (tilde expansion supported, can override with --workspace CLI flag).
  • claude: { model?, skipPermissions?, timeout? }
  • heartbeat: { enabled, intervalMs, prompt?, activeHours? }
  • gateway: { port, host?, hooks: { [type]: { token, sessionKey } } }
  • sessions: { storeDir }

Workspace layout (OpenClaw-compatible)

Files loaded (if present) into the system prompt in order: SOUL.md, AGENTS.md, USER.md, MEMORY.md, TOOLS.md, HEARTBEAT.md. Drop your OpenClaw workspace here and it will be read as-is.

Operating the daemon

• Gateway: HTTP server with /health, /hooks (typed webhooks, Bearer token auth per hook), /message (direct sessionKey with Bearer token lookup). System prompt comes from workspace build.
• Sessions: JSONL transcripts stored under sessions.storeDir; metadata tracks lastActive.
• Heartbeat: optional periodic Claude call. Default prompt expects HEARTBEAT.md and replies HEARTBEAT_OK when idle. Honors intervalMs and optional activeHours window; prevents overlapping runs.
• Tools available to the agent (internal): exec (shell with timeout), read/write (file IO). Keep commands minimal and prefer explicit cwd when using exec.

Safe upgrade practice (recommended)

• Loop: git pull (or cherry-pick) → npm run build → npm test → npm run check → restart daemon via your supervisor (e.g., systemd). Keep previous dist for rollback before swapping.

Further reading

• ABOUT — motivation and inspection guarantee.
• DESIGN — principles, architecture sketch, upgrade protocol.
• docs/CAPABILITY_INVENTORY.md, docs/GATEWAY_USAGE.md, docs/SUBSTRATE_EQUIVALENCE.md — deeper details.