TTMT Docs

Appearance

Connect with Claude Code ​

Claude Code is Anthropic's command-line assistant. Once you connect it to TTMT, you can ask plain-English questions about your own trades, performance, settings, and open positions right from your terminal — and get answers pulled live from your account.

The connection is read-only. Claude can read your data to answer your questions, but it can never place, modify, or close a trade, change a setting, or move funds. There is no write access, and none is coming through this connector.

Availability

MCP access depends on account enablement and the subscription-access rollout, separately from Developer API access. See Account access. If sign-in or consent fails, contact support to check your access.

Add the connector ​

Run this once in your terminal:

bash
claude mcp add --transport http ttmt https://app.telegramtometatrader.com/api/mcp/mcp

That registers a connector named ttmt pointing at the TTMT MCP server. The --transport http flag is required — TTMT uses the Streamable HTTP transport, and the older SSE transport is turned off.

You can name the connector anything you like in place of ttmt; the name is just how you'll refer to it inside Claude Code.

Sign in the first time you use it ​

The command above only registers the connector — it does not sign you in yet. The first time you actually ask a question that needs your TTMT data, Claude Code opens a browser sign-in for you.

Here's the whole flow:

  1. You ask a question that touches your TTMT data.
  2. Claude Code opens a sign-in page in your browser.
  3. You sign in with your TTMT account — the same login you use for the dashboard. There is no separate MCP password or API key to manage.
  4. You review a consent screen and approve read-only access (the ttmt:read scope).
  5. Done. Claude Code can now answer questions from your account.

Every request is re-verified against your login — there are no long-lived sessions sitting around. If you want to cut Claude off, remove the connector from Claude Code and the access is gone.

Why read-only, always

The consent screen only ever asks for read access. There is no write permission to approve, because no tool on this connector can act on your account. Even if a question is phrased like a command — "close my losing trades" — Claude can only look; it cannot execute.

Ask your first question ​

Once you're connected, just ask in natural language. For example:

You: Give me my morning trading summary.

Claude pulls a one-call rollup of your day — closed profit and loss, trades opened, positions still open, any halted accounts, signals that were skipped, and your best and worst channel — then reads it back to you in plain English.

A few more things worth asking:

  • "Which of my accounts are halted right now?"
  • "Show me my losing XAUUSD trades this week."
  • "What's my total floating P&L right now?"
  • "How does my Aggressive profile compare to Balanced last month?"
  • "Why didn't my last EURUSD signal get traded?"
  • "If I got a XAUUSD BUY signal now, what would my order plan look like?"

You don't need to know tool names or exact filters — describe what you want and Claude picks the right read for you. For the full list of what it can answer, see the tool reference.

If a question fails ​

  • The connection won't authorize. Check the URL, consent and recent request volume, then message support to check your MCP access.
  • Answers suddenly start failing after a burst of questions. The connector allows 60 requests per minute, shared across every client you've connected. If you go over, requests can briefly look like a sign-in failure rather than a clear rate-limit message — a known rough edge. Wait a minute and try again.