Agent API

This sections documents the Agent API

The Agent API is currently in Beta - Please report any issues or feedback in our Telegram Group

---

About

The Agent API is the next evolution of Katoshi - Allowing you to send complex commands to your AI Agents using natural language via simple POST requests or using webhooks. Read more about agent capabilities here:

Agents

---

Quick Start

POST https://api.katoshi.ai/agent?id=USER_ID

Replace USER_ID with your unique ID, which can be found in the Settings Page

Headers

Name	Type	Description
Content-Type*	String	application/json

Request Body

Name	Type	Description
api_key*	String	Katoshi API Key
agent_id*	Number	ID of your agent
message*	String	Message sent to your agent
task_id	Number	ID of a task

The message field is required if no task_id is specified

Example Request

{
  "message": "Long BTC with $11 if the 15m RSI is oversold",
  "api_key": "12345678-1234-1234-1234-123456789012",
  "agent_id": 7
}

For TradingView, add the Request JSON directly in the alert message field.

---

task_id

Passing a task_id parameter allows you to send predefined messages to your agent.

• Create tasks in your agent settings first, then reference the task_id in your request.

Example Scenario

Task message

Open a long position with $50 size, 1% stop loss, and 3% take profit when the following conditions are met:
- The 15m RSI is oversold
- The funding rates is negative
- There is no other open perpetual positions in my account

Request

{
  "message": "Apply this strategy for BTC:",
  "api_key": "12345678-1234-1234-1234-123456789012",
  "agent_id": 7,
  "task_id": 2
}

Final Message

Apply this strategy for BTC:

Open a long position with $50 size, 1% stop loss, and 3% take profit when the following conditions are met:
- The 15m RSI is oversold
- The funding rates is negative
- There is no other open perpetual positions in my account

The message parameter is optional when using task_id; but can be included to append it to the task message to create dynamic reuests.

---

Testing

Test your commands using the Agents chat interface in Katoshi.

Results can vary tremendously based on your selected LLM model, system prompt, and message sent.

Agents