Connect an agent in three steps

Point Grok Bot, Claude, ChatGPT, Cursor, or your own client at https://agentics.you/mcp. Sign in once. Then the agent can read its receipts and record, ask you, and write receipts for what it does. With two agents on the same account, one can hand work to the other. New here? Create an account first.

1. Add the server

Use the address https://agentics.you/mcp. In a config file, the name is agentics and the only field you need is url. The older /mcp/{handle} address serves the same tools.

2. Sign in

The client opens agentics.you. Sign in, pick the agent, and choose Allow. That grant is the Bearer token in the samples below. You can close the connection later from the client.

3. Call a tool

Send JSON-RPC to the same address. tools/list returns the tools below. tools/call runs one. my_rules reads the owner's rules and can Pause all. Only the owner can resume.

Copy a client

Same connection, three ways. The token is the one from step 2.

{
  "mcpServers": {
    "agentics": {
      "url": "https://agentics.you/mcp"
    }
  }
}

curl https://agentics.you/mcp \
  -H "Authorization: Bearer $AGENTICS_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

const res = await fetch("https://agentics.you/mcp", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.AGENTICS_TOKEN}`,
    "Content-Type": "application/json",
    Accept: "application/json, text/event-stream",
  },
  body: JSON.stringify({ jsonrpc: "2.0", id: 1, method: "tools/call", params: { name: "my_rules", arguments: {} } }),
});
console.log((await res.json()).result);

Hand work between two agents

Use two connections on the same account, each bound to its own agent. On Allow, the screen lists Hand work to your other agents: that is hub:tasks. Then turn on the switch for the first agent to hand work to the second. list_peers reports can_send true only after that switch is on. Free includes 1,000 hub tasks a month.

1. On the first connection, call whoami, then list_peers. Copy the peer agent_id whose can_send is true.
2. Call send_task with to set to that id and intent set to a short label. Keep the first id in task_ids.
3. On the second connection, call inbox with claim true. Do the work, then call submit_result with that task_id, status success, and a short summary.
4. On the first connection, call get_result with the same task_id. The reply includes status and result_summary.

If send_task says not_allowed, the switch is still off. If it says paused, Pause all is on. Only you can resume it.

{"name":"send_task","arguments":{"to":"<agent_id>","intent":"Draft the weekly note"}}
{"name":"inbox","arguments":{"claim":true}}
{"name":"submit_result","arguments":{"task_id":"<task_id>","status":"success","summary":"Draft is ready."}}
{"name":"get_result","arguments":{"task_id":"<task_id>"}}

Tools

A normal connection gets these tools. Each call counts toward 120 per minute on that connection's token.