CLI reference
clawmetry#
With no subcommand, clawmetry runs the dashboard.
clawmetry # foreground, port 8900
clawmetry --port 9000 # a different port
clawmetry start # background serviceService subcommands#
| Command | Effect |
|---|---|
clawmetry start | Start as a background service |
clawmetry stop | Stop the background service |
clawmetry restart | Restart it |
clawmetry status | Service status, port and uptime |
clawmetry uninstall | Remove the background service |
clawmetry help | Show help |
Two `status` commands
clawmetry status reaches the management CLI's status — daemon state, cloud
sync, store size. The dashboard's own service status is what you get from the
service subcommand context. In practice, run clawmetry status; it is the one
you want. → status
Flags#
These apply to clawmetry itself and to every service subcommand.
Server#
| Flag | Default | Effect |
|---|---|---|
--port, -p | 8900 | Port to listen on |
--host, -H | 127.0.0.1 | Bind address. 0.0.0.0 exposes it to your network — read Ports and networking first |
--version, -v | Print the version and exit | |
--no-debug | Disable debug mode and auto-reload |
Directories#
| Flag | Effect |
|---|---|
--workspace, -w | Agent workspace directory |
--data-dir, -d | OpenClaw data directory, e.g. ~/.openclaw |
--openclaw-dir | OpenClaw config directory. Env: CLAWMETRY_OPENCLAW_DIR |
--log-dir, -l | Log directory |
--sessions-dir, -s | Sessions directory holding transcript .jsonl files |
--metrics-file, -m | Path to the metrics persistence JSON file |
Auto-detection covers all of these in a normal install. Set them when a runtime lives somewhere non-standard and you would rather not use the environment variables.
Display#
| Flag | Effect |
|---|---|
--name, -n | Your name, shown in the Flow tab |
Streaming and limits#
| Flag | Default | Effect |
|---|---|---|
--sse-max-seconds | — | Maximum lifetime of a Server-Sent Events connection |
--max-log-stream-clients | 10 | Concurrent clients on the log stream |
--max-health-stream-clients | 10 | Concurrent clients on the health stream |
The stream caps exist because SSE connections are cheap to open and not free to hold. Raise them on a machine that several people watch at once.
Monitoring and fleet#
| Flag | Effect |
|---|---|
--monitor-service NAME:PORT | Add a service to monitor. Repeatable. |
--mc-url | Management/control URL |
--fleet-api-key | Fleet authentication key. Env: CLAWMETRY_FLEET_KEY |
--fleet-db | Fleet database path. Env: CLAWMETRY_FLEET_DB |
Examples#
# A second dashboard against a different workspace, for testing
clawmetry --port 8901 --workspace ~/agents/staging --no-debug
# Watch two services alongside the agents
clawmetry --monitor-service gateway:18789 --monitor-service proxy:4100
# Run on a LAN address, behind something that terminates TLS
clawmetry --host 127.0.0.1 --port 8900 # then reverse-proxy to itBinding to 0.0.0.0
The dashboard reads your agents' transcripts. Exposing it on a routable
interface exposes those. If you need remote access, put it behind an
authenticating reverse proxy and set CLAWMETRY_API_TOKENS. See
Ports and networking.