CLI

The single binary is clickclack. Source: apps/api/cmd/clickclack/main.go. It can host a server and act as a scriptable local or remote chat client. The broader agent-friendly contract is documented in agent-friendly-cli.md.

clickclack <command> [flags]

Commands:
  serve      run the HTTP/WebSocket server (default if no command given)
  migrate    apply embedded SQL migrations
  admin      bootstrap, FakeCo seed, user create, invite create, bot create, events prune, magic-link create
  backup     write a SQLite backup file
  export     write a JSON dump to a file or stdout
  login      consume a magic-link token and store/print a session token
  logout     remove stored client credentials
  whoami     print the current server-side user
  status     print selected server/user/workspace/channel
  canary     prove a human ClickClack -> OpenClaw -> quoted bot reply round trip
  workspaces list
  channels list
  send
  messages send
  messages list
  threads open
  threads reply
  reactions add
  reactions remove

Client commands accept these common flags before the command or on the command itself:

Flag	Env	Default	Notes
`--server`	`CLICKCLACK_SERVER`	`http://localhost:8080`	Remote server URL.
`--token`	`CLICKCLACK_TOKEN`	stored config	Sends `Authorization: Bearer`.
`--user`, `--user-id`	`CLICKCLACK_USER_ID`	unset	Sends `X-ClickClack-User`; local development/test escape hatch.
`--workspace`	`CLICKCLACK_WORKSPACE`	first visible workspace	ID, slug, or name.
`--channel`	`CLICKCLACK_CHANNEL`	`general`, then first visible channel	ID or name. Channel IDs are resolved across visible workspaces unless `--workspace` is set.
`--json`	—	false	Machine-readable JSON output.
`--plain`	—	false	Stable single-field output, usually an ID or token.
`--no-input`	—	false	Reserved for non-interactive flows.

Stored client config lives at ~/.config/clickclack/config.json. Stored token, workspace, and channel defaults are scoped to their saved server URL. If --server or CLICKCLACK_SERVER points somewhere else and no explicit --token is supplied, the CLI will not reuse the saved bearer token. It also will not reuse the saved workspace or channel unless the selected server matches.

If --user or CLICKCLACK_USER_ID is set and no explicit --token is supplied, the CLI also skips the stored bearer token so the server can honor the requested dev user. If both --token and --user are supplied, the server's auth order applies and the bearer token wins.

#`serve`

clickclack serve \
  --addr :8080 \
  --data ./data \
  --db sqlite://./data/clickclack.db \
  --config ./clickclack.json \
  --dev-bootstrap=true

Loads config from --config (JSON) layered on top of env vars (CLICKCLACK_*).

CLI flags win when explicitly set. See configuration.md.

Creates <data>, <data>/uploads, <data>/logs.
Opens SQLite (modernc) with WAL, foreign keys, and busy_timeout=5000,

then runs migrations.

When --dev-bootstrap=true, creates a Local Captain user and a

ClickClack workspace if the DB is empty. The default is false.

Logs the resolved listen URL and the dev-auth user ID.
--environment labels opt-in metrics; --metrics-enabled=true exposes them

at /metrics.

#`migrate`

clickclack migrate --data ./data --db sqlite://./data/clickclack.db

Idempotent: each migration in apps/api/internal/store/sqlite/migrations/ is recorded in schema_migrations and skipped on subsequent runs. Use this in deployments before flipping traffic to a new build.

#`admin`

#`admin bootstrap`

clickclack admin bootstrap --name "Peter" --email steipete@gmail.com

Creates the first user, workspace, and general channel if none exist. Idempotent — re-running prints the existing user's ID. Output is the usr_... ID on stdout, ready to capture in a shell script.

#`admin user create`

clickclack admin user create --name "Ari" --email ari@example.com [--workspace wsp_...]

Creates a user. With --workspace, also adds them to that workspace as a member. Prints the new user ID.

#`admin fakeco seed`

CLICKCLACK_ENVIRONMENT=fakeco clickclack admin fakeco seed --data ./data

Creates the deterministic synthetic FakeCo workspace, users, channels, roots, and thread replies. Reruns reuse fixed identity subjects and message nonces. The exact environment confirmation prevents accidental use against an unlabelled or production deployment. The JSON output is a non-secret resource manifest. See fakeco.md.

#`admin invite create`

clickclack admin invite create --workspace wsp_...

Mints an invite token (created by the first user in the DB, which is the owner in single-tenant deployments). Prints the token. There is no consume endpoint over HTTP yet — invite tokens are reserved for V1 work.

#`admin bot create`

clickclack admin bot create \
  --workspace wsp_... \
  --created-by usr_manager \
  --name "OpenClaw Service" \
  --handle openclaw \
  --scopes bot:write \
  --plain

clickclack admin bot create \
  --workspace wsp_... \
  --owner usr_peter \
  --created-by usr_peter \
  --name "Peter's OpenClaw" \
  --handle peter-openclaw \
  --scopes bot:write

Creates a kind=bot user, adds it to the workspace, and mints a scoped ccb_... bot token. --owner makes it a user-owned bot; omitting --owner makes it an independent service bot. Plain output prints only the raw token. For a service bot, --created-by must identify a workspace owner or moderator. For a user-owned bot, it must match --owner. JSON output includes {bot, bot_token, token}. See features/bots.md and bot-installs.md.

#`admin magic-link create`

clickclack admin magic-link create --email steipete@gmail.com --name "Peter"

Mints a magic-link token. Hand it to the user; they POST it to /api/auth/magic/consume to get a session. See features/auth.md.

#`admin events prune`

clickclack admin events prune \
  --workspace wsp_... \
  --older-than-days 30 \
  --keep-latest 10000

Deletes old durable realtime events for one workspace. At least one retention bound is required: --older-than-days, --before RFC3339, or --keep-latest. --keep-latest preserves the newest N events by cursor even when they are older than the cutoff. Private event recipient rows are deleted by cascade with their events.

#`backup`

clickclack backup --data ./data --out ./data/backup.db

Uses SQLite's online backup API to write a hot copy of the database. Safe to run while serve is up. The destination must be on the same filesystem as <data> if you want a fast atomic move afterwards.

#`export`

clickclack export --data ./data --out ./data/export.json
clickclack export --out -                # stdout

Writes a JSON dump of users, workspaces, channels, messages, threads, reactions, uploads metadata, and DMs. Useful for migrations between SQLite files or for one-off audits.

#Client auth

clickclack --server http://localhost:8080 login \
  --magic-token mgt_... \
  --plain

Consumes a magic-link token through /api/auth/magic/consume. By default it stores the returned session token in the client config file. Use --no-store for CI or tests that only need stdout.

logout removes the stored client config.

For remote or hosted agents, prefer login plus stored credentials or CLICKCLACK_TOKEN. Reserve --user for local/dev servers where explicit user impersonation is acceptable.

#Client reads

clickclack --server http://localhost:8080 --token sst_... whoami
clickclack --server http://localhost:8080 --token sst_... workspaces list
clickclack --server http://localhost:8080 --token sst_... channels list --workspace clickclack
clickclack --server http://localhost:8080 --token sst_... messages list --channel general --json
clickclack --server http://localhost:8080 --token sst_... threads open msg_... --json

workspaces list prints id slug name in human mode. channels list prints id name kind. messages list prints seq id author body.

#Client writes

clickclack --server http://localhost:8080 --token sst_... send --channel general "hello"
printf 'long body\n' | clickclack --server http://localhost:8080 --token sst_... send --channel general --stdin
clickclack --server http://localhost:8080 --token sst_... threads reply msg_... --stdin <reply.md

send and threads reply accept the body from a positional argument, --body, --file, or --stdin. Both also accept --reply-to msg_... to inline-quote an existing message in the same channel/thread (see features/replies.md). --plain prints only the created message ID; --json prints the API response.

Reaction commands use the existing public reaction API:

clickclack reactions add msg_01kr... "👍"
clickclack reactions remove msg_01kr... "👍"

Human output confirms the action without echoing the reaction value. --json prints the API's {event,reactions} response. --plain is unsupported and fails before making a request.

#`canary`

CLICKCLACK_TOKEN=... \
OPENCLAW_GATEWAY_HEALTH_URL=http://openclaw.internal:18789/healthz \
clickclack canary --workspace fakeco --channel e2e-canary --json

Requires a human session token. It optionally preflights the OpenClaw health URL, posts a unique correlated prompt, and waits for a bot reply quoting the request and carrying the same marker. --timeout, --poll-interval, and --correlation-id are available for controlled tests. Optional --run-id appears in JSON evidence without changing the request; case_id is the request message ID. Never pass credentials inside the health URL.

#Exit codes

0 on success, non-zero on any error (a bare log.Fatal in main). Scripts should rely on the exit code, not the log line format.

CLI

#serve

#migrate

#admin

#admin bootstrap

#admin user create

#admin fakeco seed

#admin invite create

#admin bot create

#admin magic-link create

#admin events prune

#backup

#export