Skip to content

Connect your AI assistant

Backtest360 exposes a Model Context Protocol (MCP) server, so an AI assistant or agent can use the engine's tools directly — discover indicators, build and validate strategies, run backtests, and read the results — in a single conversation.

It's the same engine and the same API key as the REST API and Python client. The MCP server contains no AI of its own; it's a thin adapter that forwards your requests to the engine and shapes the responses for the assistant.

Two ways to connect

Point your MCP client at the hosted endpoint over HTTP and send your key in an X-API-Key header:

  • URL: https://mcp.backtest360.com/mcp
  • Header: X-API-Key: b360_...

Most MCP clients accept a configuration like this:

{
  "mcpServers": {
    "backtest360": {
      "type": "streamable-http",
      "url": "https://mcp.backtest360.com/mcp",
      "headers": {
        "X-API-Key": "b360_..."
      }
    }
  }
}

Add that to your client's MCP configuration, restart the client, and the Backtest360 tools become available. Some MCP clients also offer a CLI to register an HTTP server in a single step — check your client's documentation for the exact command.

Local

Prefer to run the server yourself? Install it and let your MCP client launch it over stdio — most desktop MCP clients support this through their JSON configuration file:

pip install backtest360-mcp

{
  "mcpServers": {
    "backtest360": {
      "command": "backtest360-mcp",
      "env": {
        "BACKTEST360_API_KEY": "b360_..."
      }
    }
  }
}

BACKTEST360_API_KEY is the only required variable. The rest are optional:

Variable Default Purpose
BACKTEST360_ENGINE_URL https://api.backtest360.com The engine the server forwards to.
BACKTEST360_MCP_TIMEOUT 300 Per-request timeout, in seconds.
BACKTEST360_MCP_MAX_OUTPUT_BYTES 100000 Cap on the size of a tool response.

The local server talks to the same engine; your key governs access either way.

Authentication

The MCP server uses the same b360_ key as everything else. It never stores the key — it forwards it per request, and the engine enforces your plan's permissions, rate limits, and quotas on every call. See Authentication to get a key.

What the assistant can do

The server exposes the engine as a set of tools — see Tools for the full list. A typical session: discover what exists, build a strategy, validate it, run a backtest, read the numbers — the assistant chains the tools for you.

All numbers come from the engine. The server never computes or estimates results; it only forwards your requests and shapes what the engine returns.

Next

Browse the tools →