Task Management¶
CodeTether provides comprehensive task management for AI agent workflows.
Task Lifecycle¶
stateDiagram-v2
[*] --> Pending: Create
Pending --> Running: Assigned
Running --> Completed: Success
Running --> Failed: Error
Pending --> Cancelled: Cancel
Running --> Cancelled: Cancel
Completed --> [*]
Failed --> [*]
Cancelled --> [*]
Task States¶
| State | Description |
|---|---|
pending |
Task created, waiting for worker |
running |
Worker is executing the task |
completed |
Task finished successfully |
failed |
Task encountered an error |
cancelled |
Task was cancelled |
Creating Tasks¶
Via A2A Protocol¶
{
"jsonrpc": "2.0",
"method": "message/send",
"params": {
"message": {
"parts": [{"type": "text", "content": "Analyze this codebase"}]
}
},
"id": "1"
}
Via CodeTether API¶
curl -X POST /v1/agent/codebases/{id}/tasks \
-d '{"title": "Add tests", "prompt": "Add unit tests"}'
Task Priority¶
Tasks can be prioritized (higher number = higher priority):
Task Routing¶
Tasks are routed to workers based on codebase_id:
| Codebase ID | Routing Behavior |
|---|---|
{specific_id} |
Routed only to workers with that codebase registered |
global |
Routed to all workers with a global codebase |
__pending__ |
Registration tasks that any worker can claim |
Global Tasks¶
Tasks with codebase_id: "global" are sent to all workers that have registered a global codebase. This is useful for:
- Cross-codebase operations
- Administrative tasks
- Tasks that don't belong to a specific project
# Create a global task via MCP
curl -X POST /v1/agent/tasks \
-d '{
"title": "System maintenance",
"prompt": "Run system health checks",
"codebase_id": "global"
}'