The trading journal MCP server — ask Claude or ChatGPT about your own trades
Connect it once and your own Claude, Claude Code or ChatGPT can read your real trade history: every fill, every stat, your tags and your diary. It reads, it never writes, and it only ever sees your account.
Built on the journal 170,000+ traders already use.
$ claude mcp add --transport http tmm https://tradermake.money/api/v2/mcpEndpoint:https://tradermake.money/api/v2/mcp
Works with Claude, Claude Code and ChatGPT
What is the TMM MCP server?
It gives Claude, Claude Code and ChatGPT read access to your own crypto trading history: your trades, your PnL, your win rate, your risk numbers, your diary. An AI trading journal that answers with your numbers instead of general advice.
MCP stands for Model Context Protocol. It is the open standard AI clients use to reach tools and data that live outside the chat. Connect the TMM server once — sign in and approve, or paste a personal token — and your assistant stops guessing from screenshots you paste in. It reads the trades themselves.
We host it, so there is nothing to install and nothing running on your machine. It reads the same journal the web app shows you, across every exchange you have connected, and it cannot change a single thing in it.
MCP endpoint:https://tradermake.money/api/v2/mcp
What it looks like in the chat
You never learn a command or pick a tool. You ask the way you would ask someone who has read your journal, and the assistant goes and gets the numbers itself. These are the questions traders actually ask it, and the kind of answer that comes back.
How did last month go?
84 trades, 61% of them green, +$3,180 net after fees. Your average winner is $132 against an $88 average loser, so your payoff is 1.5. BTC longs carried the month at +$3,590; ETH shorts handed $410 of it back.
The answers above are examples and the numbers in them are invented. Yours are read live from your own trades every time you ask.
What your AI can read: trades, PnL, win rate, risk, diary
Behind those answers sit 15 tools, and every one of them is read-only. You never call them by name — your assistant picks whichever one fits the question. Open the list if you want to see exactly what it can reach.
Connect in two minutes
Paste one URL, sign in with the TMM account you already have, and approve what your assistant may read. There is nothing to copy and nothing else to fill in — your client handles the sign-in itself.
- 01.
Paste the endpoint
In your client's connector settings, add a custom connector and paste the TMM endpoint. Copy it exactly as it is written on this page: https, no trailing slash, nothing after mcp.
- 02.
Sign in and approve
Your client opens TMM in the browser. Log in the way you normally do, choose whether the assistant may also read your trading diary, and approve. The connection belongs to your account and lives in your settings.
- 03.
Ask a question
Start a chat and ask something about your trading. Your client picks up the 15 read-only tools on its own and calls whichever one fits the question, so you never have to name a tool.
Claude and Claude Desktop
The same steps on claude.ai and in the desktop app — Claude Desktop syncs connectors from the web, so you only set it up once.
- 1Open Settings and go to Connectors.
- 2Choose Add custom connector.
- 3Paste the TMM endpoint into the URL field.
- 4Press Add. That is the whole setup — there is nothing else to fill in.
- 5Press Connect and approve the permissions on the TMM screen that opens.
The Free plan allows one custom connector, so you may have to remove another one first. Paid plans have room for several.
Claude Code
Register the server once from any terminal, then sign in. The command is global, so every project can reach it.
claude mcp add --transport http tmm https://tradermake.money/api/v2/mcp
claude mcp login tmm
You can also just type /mcp inside Claude Code and pick TMM to sign in. Signing in opens your browser; over SSH or in a container add --no-browser and Claude Code prints a URL for you to open on your own machine instead.
ChatGPT
ChatGPT reaches custom MCP servers through developer mode, which is not available on every plan — check what your account includes.
- 1In Settings, open Connectors and turn on developer mode.
- 2Choose to create a new connector.
- 3Paste the TMM endpoint as the MCP server URL and pick OAuth as the authentication.
- 4Approve the permissions on the TMM screen when the browser opens.
If ChatGPT will not connect on its own, write to support. We will send you the one setting to paste into the connector's advanced options — everything else on the form stays as it is.
Cursor and other MCP clients
Most other AI apps that support connectors will work too. Give them the same address: they will either open the TMM sign-in themselves, or ask you for a personal token.
Three things that break a connector
- Copy the URL exactly
The address you type is part of what gets approved. A different host, an extra path segment or a stray character and the sign-in is refused, because the resource no longer matches the one you asked for.
- HTTPS, and no trailing slash
It ends at mcp, with no trailing slash and no extra path. Plain http will not connect at all.
- The diary is a separate tick
The consent screen asks for your trades and your diary as two separate permissions. Leave the diary unticked and the diary tool is not handed to the assistant at all — it will tell you it cannot see your notes. Re-approve from Settings if you change your mind.
What you are approving
The consent screen asks for two things separately, and both are read-only. Nothing on it lets an assistant place an order, move money or change your journal.
- Your trades and stats — trades and fills, PnL, win rate, risk numbers, tags and exit reasons.
- Your trading diary — the notes you wrote yourself. Leave it unapproved and the assistant simply cannot read them.
The connection is scoped to your own account and can see nobody else's data. Revoke it from your TMM settings at any time and access stops immediately.
Prefer a static token?
The personal access token is the simpler option for CI, for scripts, and for Claude Code on a machine with no browser to open. Mint one in the app under Settings, Connect your AI.
claude mcp add --transport http tmm https://tradermake.money/api/v2/mcp \ --header "Authorization: Bearer <your-token>"
Header: Authorization: Bearer <your-token>
Read-only by design
An AI that can move money is a much riskier thing to plug into your account. This one has no way to do it. That comes from how the server was built, not from a setting you have to remember to turn on.
Read-only tools only
Your assistant is only ever offered tools that read. Nothing that could add, change or delete anything in your journal is available to it.
No orders, no withdrawals
There is no order call and no withdrawal call behind any of the tools. The server cannot trade for you, and it never sees your exchange API keys.
Your data only
The connection is tied to your account. There is no way for it to read anyone else's data, whatever the assistant asks for.
Revoke or rotate any time
Connections and tokens both sit in Settings with your other credentials. Revoke a connection or delete a token and access stops straight away. Connect again whenever you want.
Rate limited
Calls are rate limited per token, separately from the rest of your account, so a runaway agent loop burns through its own budget and nothing else.
The same limits as the app
The MCP server runs on the app's own data layer, so your plan's history window applies exactly as it does on the web. No back door, no extra exposure.
Plans and data depth
The MCP server itself is free. Your plan only decides how far back your AI can read, and that is the same history window you already have in the web app.
Any verified user can connect, whatever plan they are on. When the plan cuts the history off, the model is told to say so rather than fill the gap with a guess.
Compare plansTMM MCP vs other trading MCP servers
Trading MCP servers come in three kinds: brokerage connectors that can execute, market-data feeds, and journals like this one. They answer very different questions, so it is worth knowing which kind you are plugging in.
Competitor details come from their public documentation as of July 2026 and can change, so check their own pages before you decide.