WorkClaw/Documentation
Sign inGet started

API Endpoints

Complete reference for WorkClaw REST API endpoints. Manage Claws, deployments, skills, team members, and chat programmatically.

How are endpoints organized?

All endpoints are prefixed with /v1 and grouped by resource. Requests and responses use JSON. Authentication is required for all endpoints -- see API Authentication.

What Claw endpoints are available?

MethodPathDescriptionScope
GET/v1/clawsList all Claws in your workspaceclaws:read
GET/v1/claws/:idGet a specific Claw's configurationclaws:read
POST/v1/clawsCreate a new Clawclaws:write
PATCH/v1/claws/:idUpdate a Claw's configurationclaws:write
DELETE/v1/claws/:idDelete a Clawclaws:write

Claw objects include identity, personality, installed skills, active connections, and current deployment status.

What deployment endpoints are available?

MethodPathDescriptionScope
GET/v1/deploymentsList recent deploymentsdeployments:read
GET/v1/deployments/:idGet deployment details and statusdeployments:read
POST/v1/deploymentsTrigger a new deploymentdeployments:write
POST/v1/deployments/:id/rollbackRoll back a deploymentdeployments:write

When triggering a deployment, you can optionally specify a deploy strategy to override the team default.

What skill endpoints are available?

MethodPathDescriptionScope
GET/v1/claws/:id/skillsList installed skillsskills:read
POST/v1/claws/:id/skillsInstall a skillskills:write
DELETE/v1/claws/:id/skills/:skillIdRemove a skillskills:write

What team endpoints are available?

MethodPathDescriptionScope
GET/v1/team/membersList team membersteam:read
POST/v1/team/membersInvite a new memberteam:write
PATCH/v1/team/members/:idUpdate a member's roleteam:write
DELETE/v1/team/members/:idRemove a memberteam:write

What chat endpoints are available?

MethodPathDescriptionScope
POST/v1/chat/:clawId/messagesSend a message to a Clawchat:write
GET/v1/chat/:clawId/messagesRetrieve conversation historychat:read

The chat endpoint returns the Claw's response synchronously. For long-running tasks, the response includes a taskId you can poll via GET /v1/tasks/:taskId. See Tasks for more on task management.

How are errors returned?

All errors follow a consistent format:

{
  "error": {
    "code": "not_found",
    "message": "Claw not found",
    "status": 404
  }
}

Common error codes: unauthorized (401), forbidden (403), not_found (404), rate_limited (429), internal_error (500).

Related documentation