Commands and flags

The app

clauder                      # listens on the loopback only, port 4221
clauder --host 0.0.0.0       # open access over the network
clauder --port 4300          # another port
clauder --print-config       # where the settings and sessions are, and what the access key is
clauder --install-agent      # bring the server up in the background at login

--print-config is the first thing to run when something doesn't add up: it prints both the path to config.json and the session directories the server actually reads.

Docker

docker compose up -d                      # bring it up
docker compose up -d --build api          # after editing api/Dockerfile
docker logs clauder-api --since 5m        # api logs
curl http://localhost:4221/api/health     # health

The rebuild gotcha

docker compose build api builds the image, but the running container stays on the old layer until docker compose up -d api. Better to check by fact:

docker exec clauder-api sh -c 'command -v java gh curl'

api and web run with hot-reload: changes in api/src and web/app are picked up without a rebuild. The api, however, comes back not immediately but 15 seconds after a successful build — otherwise every saved edit would kill live agent sessions.

Building

./scripts/build-local.sh --check          # what this machine is missing
./scripts/build-local.sh                  # build everything for your own system
./scripts/build-desktop.sh                # the server only, for the current platform
./scripts/build-desktop.sh all            # windows + macos (arm64/x64) + linux
cd desktop && cargo tauri build           # the windowed shell (your own OS only)
cd android && ./build.sh                  # a signed release APK

Flags for build-local.sh: --server (server only), --shell (window only), --skip-web (don't rebuild the web UI).

Checks and releases

cd api && npx tsc --noEmit                # typecheck the api
cd web && npx nuxt typecheck              # typecheck the web UI
./scripts/release.sh [vX.Y.Z]             # tag, APK, desktop, gh release
./scripts/detect-claude-homes.sh          # find the ~/.claude profiles on the machine
./scripts/hostrun '<command>'             # a shell on the host through the ssh bridge

Composer slash commands

These work in a chat, not in a shell:

Command	What it does
/model, /effort, /permissions, /ultracode, /workflows	change the live session's settings
/context, /cost	show its state
/btw	a side question in the same conversation
/continue	"Continue" and send right away
/agents, /chats, /files, /search	navigate the interface
/login, /logout	sign in to Claude with an OAuth-link window, and sign out

A line starting with ! doesn't go to the model at all: it is run by the shell in this chat's directory (!git status, !npm test), and the output lands in the feed.

An unknown name doesn't count as a command and goes out as a message — that is how the CLI's own commands (/compact, /init, your project ones) pass through.

Bot commands

/chats, /new, /where, /agents, /status, /stop, /app — see The Telegram bot.