Cloud-based REST APIs for managing the Kubiya platform resources
The Kubiya Platform APIs are cloud-based REST APIs that allow you to programmatically manage your Kubiya resources including agents (teammates), sources (tool collections), knowledge entries, tasks, and runners.
These APIs are different from the Agent Server APIs which are exposed by orchestration servers for direct agent interactions.
All Platform APIs use the following base URL:
All Platform API endpoints require authentication using a UserKey:
Execute and manage workflow tasks across your organization
Key endpoints:
GET /api/v1/tasks
- List all tasksPOST /api/v1/tasks
- Create a new taskPUT /api/v1/tasks/{taskId}/cancel
- Cancel running tasksGET /api/v1/tasks/{taskId}/logs
- View task execution logsDeploy and manage Kubernetes operators that execute workflows
Key endpoints:
GET /api/v3/runners
- List all runnersPOST /api/v3/runners/{runner}
- Restart a runnerGET /api/v3/runners/{runner}/health
- Check runner healthPOST /api/v3/runners/{runner}/helm
- Manage Helm integrationsCreate and manage AI teammates with custom tools and knowledge
Key endpoints:
GET /api/v1/agents
- List all agentsPOST /api/v1/agents
- Create a new agentPUT /api/v1/agents/{agentId}
- Update agent configurationGET /api/v1/agents/{agentId}/integrations
- Manage integrationsManage tool sources and repositories attached to agents
Key endpoints:
GET /api/v1/sources
- List all sourcesPOST /api/v1/sources
- Create a new sourcePUT /api/v1/sources/{sourceId}
- Sync source with repositoryGET /api/v1/sources/load
- Discover and load external sourcesStore and retrieve contextual information for enhanced agent performance
Key endpoints:
GET /api/v1/knowledge
- List all knowledge entriesPOST /api/v1/knowledge
- Create knowledge entryGET /api/v1/knowledge/search
- Search knowledge baseGET /api/v1/knowledge/{id}/versions
- View version historyAll successful API responses return appropriate HTTP status codes:
200
- Successful GET, PUT, DELETE operations201
- Successful POST operations (resource created)All APIs follow a consistent error response format:
Common HTTP error codes:
400
- Bad Request (invalid parameters)401
- Unauthorized (invalid API key)403
- Forbidden (insufficient permissions)404
- Not Found (resource doesn’t exist)500
- Internal Server ErrorList endpoints that return large datasets use consistent pagination:
Here’s a complete example of creating and executing a task:
The Platform APIs have the following rate limits:
Operation Type | Limit |
---|---|
Read operations (GET) | 1000 requests/minute |
Write operations (POST/PUT) | 200 requests/minute |
Delete operations | 50 requests/minute |
Task creation | 100 tasks/minute |
Search operations | 300 requests/minute |
The Platform APIs use path-based versioning:
/api/v1/
- Current stable version for most resources/api/v2/
- Enhanced version for specific resources (agents)/api/v3/
- Latest version for runnersWhen breaking changes are introduced, a new version path is created while maintaining backward compatibility for existing versions.
Always implement proper error handling:
Implement exponential backoff for rate-limited requests:
Handle paginated responses properly:
Keep API keys secure:
Direct orchestration server interactions
Official SDKs for Python, TypeScript, and Go
Real-world usage examples and tutorials