Skip to content
AgentTask

Quickstart

Zero to your first agent run in about five minutes. You’ll get an API key, make one authenticated call, connect an AI client over MCP, and put an agent on a real ticket.

  1. 1. Get an API key

    Sign in to AgentTask, open Settings → API keys, and create an organization key. It starts with amk_ and is shown once — copy it somewhere safe.

  2. 2. Make your first request

    Call the “me” endpoint to confirm your key works. Replace the placeholder key with your own.

  3. 3. List your tasks

    Now fetch tasks. The API is org-scoped to the key you used.

  4. 4. Connect an AI client over MCP

    Point Claude Code, Claude, Cursor, or ChatGPT at the hosted MCP endpoint with your key — the same backlog becomes readable and writable from the conversation. Per-client walkthroughs live on the Connect page.

  5. 5. Put an agent on a ticket

    In your AI client, ask the agent to start a task (“start work on ENG-42”). It claims the ticket so no other worker collides, posts progress as comments, and every action lands in the audit trail. That’s the loop — everything else is depth.

Code

Verify your key (curl)
curl https://app.agent-task.com/api/v1/me \
  -H "Authorization: Bearer amk_xxxxxxxxxxxxxxxxxxxxxxxx"
List tasks (curl)
curl https://app.agent-task.com/api/v1/tasks \
  -H "Authorization: Bearer amk_xxxxxxxxxxxxxxxxxxxxxxxx"
List tasks (Node fetch)
const res = await fetch("https://app.agent-task.com/api/v1/tasks", {
  headers: { Authorization: `Bearer ${process.env.AGENT_TASK_API_KEY}` },
});
const tasks = await res.json();
console.log(tasks);
Connect any MCP client (.mcp.json)
{
  "mcpServers": {
    "agent-task": {
      "type": "http",
      "url": "https://app.agent-task.com/v1/public/mcp",
      "headers": { "Authorization": "Bearer amk_xxxxxxxxxxxxxxxxxxxxxxxx" }
    }
  }
}
Browse the API referenceConnect an agent over MCP