Running the daemon

One long-lived process per machine, holding the connection pool, caches, write queue, and peer endpoint. It auto-starts on first use — these commands are for explicit control.

sigil daemon status
sigil daemon start [--foreground]
sigil daemon stop
sigil daemon restart
sigil daemon logs [--follow]
sigil daemon open        # open the dashboard
sigil daemon url         # print the URL with its auth token

Why it exists

A hook runs on every prompt you type. Cold-starting Node, connecting to Postgres, and loading an embedding client costs 200–400ms — noticeable on every interaction.

The daemon keeps that warm, so hooks and CLI calls are a Unix socket round trip against an already-connected process. It also gives you one writer (a serialized queue, which matters enormously for the embedded PGlite database), one shared embedding cache, and one stable peer identity for device pairing.

Files

Path
~/.sigil/sock	Unix socket — hooks, CLI, stdio MCP
~/.sigil/sigild.pid	PID file
~/.sigil/sigild.log	Append-only log
~/.sigil/heartbeat.json	Liveness
~/.sigil/.spawn.lock	Prevents concurrent spawn races
~/.sigil/.daemon-breaker.json	Circuit breaker state

All recreated on start. Safe to delete while stopped.

The circuit breaker

If the daemon fails to spawn repeatedly, a breaker trips and stops further attempts rather than retrying forever — which would turn a broken install into a fork bomb on every prompt.

When tripped, hooks fail silently and fast, your prompts still work, and sigil doctor tells you what happened.

sigil daemon logs      # what actually failed
sigil daemon start     # clears the breaker on success

The HTTP server

Setting	Default
http.enabled	true	(restart required)
http.host	127.0.0.1	Loopback only (restart)
http.port	7777	(restart)

Serves the dashboard and the HTTP MCP transport, gated by a bearer token at ~/.sigil/gui.token.

Changing http.host exposes your brain

Binding to 0.0.0.0 makes the dashboard and MCP endpoint reachable from your network. The bearer token is the only thing protecting them. If you need remote access, use an SSH tunnel or device pairing instead.

ssh -L 7777:127.0.0.1:7777 you@your-server

Running under launchd or systemd

The daemon auto-starts on demand, so a service definition is optional. It’s worth it on a server where agents run unattended and you want the daemon up before they are.

~/.config/systemd/user/sigil.service

[Unit]
Description=Sigil memory daemon
After=network.target

[Service]
Type=simple
ExecStart=%h/.sigil/bin/sigil daemon start --foreground
Restart=on-failure
RestartSec=5

[Install]
WantedBy=default.target

systemctl --user enable --now sigil

~/Library/LaunchAgents/dev.sigil.daemon.plist

<?xml version="1.0" encoding="UTF-8"?>
<plist version="1.0">
<dict>
  <key>Label</key>            <string>dev.sigil.daemon</string>
  <key>ProgramArguments</key>
  <array>
    <string>/Users/you/.sigil/bin/sigil</string>
    <string>daemon</string>
    <string>start</string>
    <string>--foreground</string>
  </array>
  <key>RunAtLoad</key>        <true/>
  <key>KeepAlive</key>        <true/>
</dict>
</plist>

launchctl load ~/Library/LaunchAgents/dev.sigil.daemon.plist

Service managers strip your PATH

Under launchd or systemd the daemon runs with a minimal environment and may not find binaries in ~/.local/bin or an nvm directory. If you use Claude Code or Codex CLI as your LLM provider, set the path explicitly:

llm.cliPath   = /Users/you/.local/bin/claude
llm.codexPath = /Users/you/.local/bin/codex

Watching it work

sigil daemon logs --follow

The dashboard’s Activity screen is better for most debugging — it streams each hook fire, search, classification, and LLM call with model and latency, filterable by agent.

The Engine screen shows worker pool state, queue depth, and database connection detail. If ingest feels slow, that’s where the answer is.

Restarting

Most configuration changes apply live — the in-memory config cache updates on every patch. Settings that need a restart are marked in Configuration and the dashboard restarts for you when you apply them.

sigil daemon restart

sigil update restarts it automatically after fast-forwarding.

Version drift

If a daemon from an older version is still running after an update, Sigil detects the mismatch and restarts it — otherwise the stale process keeps serving and the update silently does nothing.

sigil daemon status    # shows the running version

Stopping it

sigil daemon stop

Memory is not lost — it’s in your database. Hooks fail silently while it’s down, and it auto-starts again on the next CLI call or hook fire.