CLI / REPL¶

Raven ships a terminal operator client, a Claude-Code-style REPL that connects to a Raven server and drives the same actions as the web UI: run modules, browse findings, and drive the AI advisor from the terminal.

There are two binaries:

raven-server: the server (web dashboard + engine). You start this.
raven-cli: the operator REPL/CLI. Connects to a raven-server.

(In the Pro Docker image the server CLI is installed as raven, so there is no raven-server executable inside the container. The operator REPL is a separate standalone download.)

Running the server¶

raven-server start -p 8080 --no-browser

Flag	Meaning
`-p, --port`	Port to bind (default 8080)
`--no-browser`	Don't auto-open a browser
`--host`	Interface to bind. Defaults to `127.0.0.1` (single-user). `--host 0.0.0.0` binds all interfaces. Only do this on a network you control; the single-user server has no auth, the loopback bind is the protection.

Pairing (one-time)¶

raven-server auth login      # start device pairing (opens a URL + code)
raven-server auth status     # show current pairing
raven-server auth logout     # clear the saved token

In Docker, prefix with docker exec -it raven-pro … and swap raven-server for raven, since the image installs the server CLI as raven and carries no raven-server executable:

docker exec -it raven-pro raven auth login
docker exec -it raven-pro raven auth status
docker exec -it raven-pro raven auth logout

See Install & Pair for the full flow.

The REPL¶

Launch the operator client. With nothing configured it connects to http://localhost:8080, which is raven-server's own default:

raven-cli                                  # REPL against http://localhost:8080
raven-cli -s https://your-server:8080      # point it at another server
raven-cli login https://your-server:8080   # log in first (team servers), then REPL

Inside, you drive engagements, run modules, inspect findings, and invoke the AI advisor (the same operations as the web dashboard). Type help in the REPL for the command list.

Discover commands

Both binaries are self-documenting:

raven-cli --help
raven-server --help

--help reflects the exact commands in your build, so trust it over these docs if they ever disagree.

AI from the terminal¶

Once paired, the REPL's AI commands route through Chamber Door (anonymized on the way out, budget-enforced server-side). It's the same proxy the web dashboard uses. Budget and usage are visible on app.chamberdoorsecurity.com.