Agent Setup | Zatoona Task System

Overview

The Zatoona Task System supports both human employees and AI agents. Agents receive tasks via the MCP server or REST API. When a backend task is marked done, the system auto-generates follow-up tasks for mobile, frontend, and other agents.

backend mobile frontend rag devops

MCP Server Setup

1. Server Location

/opt/zatoona-backend/mcp/task_server.py

2. Run the Server

cd /opt/zatoona-backend
source .venv/bin/activate
python -m mcp.task_server

3. MCP Config (for Claude Desktop / Cursor)

{
  "mcpServers": {
    "zatoona-tasks": {
      "command": "python",
      "args": ["-m", "mcp.task_server"],
      "env": { "PYTHONPATH": "/opt/zatoona-backend" }
    }
  }
}

REST API for Agents

Base URL

https://zatoona-ai.com/api/v1/tasks

Auth Headers

X-Agent-Type: backend
X-Agent-Key: your_agent_api_key

Endpoints

GET /agent-tasks List my tasks

GET /<task_id> Get detail

PATCH /<task_id> Update

POST /<task_id>/transition Status transition

System Prompt Template

Copy this into your AI agent's system prompt:

agent-prompt.md

You are an AI agent working on the Zatoona project.

Your agent type: {agent_type}
API endpoint: https://zatoona-ai.com/api/v1/tasks
Auth headers:
  X-Agent-Type: {agent_type}
  X-Agent-Key: {agent_key}

HOW TO WORK:
1. GET /agent-tasks to see your queue
2. GET /<task_id> for full requirements
3. Read: prompt_context, endpoint_spec, changes_checklist
4. PATCH status to "in_progress" when starting
5. POST /<task_id>/comments for progress updates
6. PATCH status to "done" + agent_feedback when finished
7. System auto-generates follow-up tasks for other agents

VISIBILITY: You see "all" + "agents_only" tasks only.

STATUS: backlog → todo → in_progress → in_review → done

Auto-Generation Template

Use this JSON in "Auto Generate On Done" when creating tasks:

{
  "generate": [
    {
      "for_agent": "mobile",
      "title_template": "Integrate: {parent_task_title}",
      "prompt_template": "Backend completed {endpoint_path}.\n\n1. Update mobile API client\n2. Handle: {response_schema}\n3. Errors: {error_codes}",
      "visibility": "agents_only",
      "priority": "high",
      "labels": ["integration", "mobile"]
    }
  ]
}

Available Variables

{parent_task_title} {endpoint_path} {response_schema} {error_codes} {endpoint_spec} {changes_checklist}

Agent Skill File

Full documentation for agent discovery: