Running teams

A run executes an assembled team’s phases on the daemon. The CLI starts the run and streams status; it does not orchestrate phases itself. Vocabulary: Concepts. Agent catalog: Agents. First install: Get started.

Lifecycle

cliq init
  → cliq assemble <team-ref>
  → cliq run [-i key=value ...]
  → cliq status / cliq logs

Optional: cliq team install … if the team is not already in the local catalog.

1. Workspace

mkdir my-project && cd my-project
cliq init
# or: cliq init --name my-project

init registers the current directory as a workspace with the daemon. Re-running it is safe.

2. Assemble a team

cliq assemble @cliq/hello-world

Assemble binds a catalog team to this workspace. The daemon validates the manifest (phases, agents, DAG, nested teams).

cliq assemble @cliq/hello-world -w ~/other-project
cliq assemble @acme/release-review -w <workspace-id>

-w, --workspace takes a workspace id, name or path. If several teams are assembled in one workspace, pick one at run time with --team.

3. Provide inputs

Teams declare what they need (task text, folder URLs, tickets, …). Pass values at run time:

cliq run -i task="Write a haiku about programming"
cliq run -i message="Hello" -i ticket=PROJ-123

-i / --input is repeatable (key=value). When required inputs are missing, cliq run lists them and exits with code 3 — supply them and run again. Some demo teams need no inputs (@cliq/hello-world).

4. Run

cliq run
cliq run --name my-feature
cliq run -t @cliq/hello-world
cliq run -w ~/my-project
cliq run --no-stream          # return once the run starts; watch with status/logs

Flag	Purpose
`-n, --name`	Name for the run (generated when omitted)
`-t, --team`	Team ref when several are assembled
`-w, --workspace`	Target workspace by id, name or path
`-i, --input`	Run input (`key=value`, repeatable)
`--label`	A label recorded with the run (`key=value`, repeatable)
`--context-id`	External correlation id; reuses the active run with the same id
`--no-stream`	Return once the run starts instead of streaming its events
`--json`	Print the started run as JSON (implies `--no-stream`)

cliq run exits 0 when the run completes and 1 when it fails. The daemon:

Resolves the assembled team and inputs
Walks depends_on and invokes each phase’s agent
Stores each phase output as a run artifact and passes predecessors as upstream
For gates: runs commands, then asks for PASS / ROUTE / ESCALATE

5. Inspect

cliq status                 # current workspace / latest run
cliq status --all           # all registered workspaces
cliq status -w my-project

cliq logs
cliq logs -f                # follow
cliq logs --last 5

Status shows phase progress. Logs show orchestrator and agent output for the run.

Stop a run

cliq instance list
cliq instance stop <run_id>

Stopping cancels the run and kills its running agent process right away; active sub-team runs are cancelled too. Cancel in CliqHub does the same.

Runs in CliqHub

When the daemon is logged in to a realm, its runs sync to CliqHub. A team must be on the realm’s team list for that. If it isn’t, the run still executes on your machine, and cliq run warns:

This run won't appear in CliqHub: @acme/my-team isn't installed in this realm. It runs here and syncs once a realm admin adds the team to the realm in CliqHub.

Only that team’s runs wait; other teams keep syncing. cliq whoami lists the waiting teams. Fix it by having a realm admin add the team to the realm, or by uninstalling it (cliq team uninstall). Details: cliqd — How teams get installed.

6. Resume / rerun a phase

Re-enter an existing run from a named phase (later phases re-execute; earlier outputs are reused when available):

cliq run <run_id> <phase_name>

run_id is a run id or name. Start-only flags (--name, --team, -i, --label, …) are ignored when rerunning. Example:

cliq run bold-arch writer

Configure agents before AI teams

LLM and connector phases need settings on the daemon:

cliq setup
# or
cliq settings set agents.cursor.api_key <key>
cliq settings set agents.jira.base_url https://yoursite.atlassian.net

Check readiness:

cliq doctor
cliq agent list

Details per agent: Agents.

Multi-workspace and remote

Goal	Command / path
Run in another workspace	`cliq run -w <name-or-path>`
CI calling Hub	User token — Auth
Enroll another machine	Realm enroll token (`cliq auth token <name> --realm <realm_id>`) → `CLIQ_DAEMON_TOKEN` for its cliqd
Drive a cliqd on another host	`CLIQ_DAEMON_HOST` + `CLIQ_DAEMON_TOKEN` — CLI
Hub-dispatched run	Hub API / BFF dispatch

Local cliq run talks to the local daemon by default.

Minimal examples

No API keys

mkdir ~/cliq-hello && cd ~/cliq-hello
cliq init
cliq assemble @cliq/hello-world
cliq run
cliq status
cliq logs

Cursor AI team

cliq settings set agents.cursor.api_key <your-cursor-api-key>
mkdir ~/cliq-cursor && cd ~/cliq-cursor
cliq init
cliq assemble @cliq/hello-world-cursor-gate
cliq run -i task="Write a haiku about programming to haiku.txt"
cliq logs


Concepts	Phases, agents, gates, upstream
Agents	Built-ins and input params
CLI	Full command reference
Auth	Tokens and realms

Start

Agents

CLI

Integrations

Running teams

Lifecycle

1. Workspace

2. Assemble a team

3. Provide inputs

4. Run

5. Inspect

Stop a run

Runs in CliqHub

6. Resume / rerun a phase

Configure agents before AI teams

Multi-workspace and remote

Minimal examples

Next

​Lifecycle

​1. Workspace

​2. Assemble a team

​3. Provide inputs

​4. Run

​5. Inspect

​Stop a run

​Runs in CliqHub

​6. Resume / rerun a phase

​Configure agents before AI teams

​Multi-workspace and remote

​Minimal examples

​Next

Lifecycle

1. Workspace

2. Assemble a team

3. Provide inputs

4. Run

5. Inspect

Stop a run

Runs in CliqHub

6. Resume / rerun a phase

Configure agents before AI teams

Multi-workspace and remote

Minimal examples

Next