AI Model Context Protocol

A read-only trading journal MCP server for AI assistants

Review your trading performance directly inside your favorite AI assistants without exporting CSVs or pasting screenshots into chat windows.

Edgelog provides a remote, read-only Model Context Protocol (MCP) endpoint for compatible clients. Its current tools expose sanitized closed trades and deterministic summaries such as win rate, profit factor, expectancy, trade count, and net P&L. Client availability and authentication support vary, so verify the setup instructions for the AI product you use.

What is the Model Context Protocol (MCP) for trading?

The Model Context Protocol (MCP) is an open technical standard pioneered by Anthropic that allows local and cloud AI models to securely connect to external data sources. Instead of relying on static file uploads or fabricated answers, an MCP-connected assistant dynamically queries live, structured data when you ask a question.

For forex and cryptocurrency traders, Edgelog’s MCP server acts as a narrow data bridge. When you ask how your EUR/USD trades performed over the last month, a compatible assistant can call deterministic tools against your actual closed-trade records. The returned totals are data-derived; the assistant’s interpretation can still be wrong and should be checked before changing a trading process.

Questions you can ask your AI assistant

Once connected, you can ask targeted questions within the current closed-trade dataset:

  • Performance audits: "What were my win rate, profit factor, expectancy, and net P&L over the last 30 days?"
  • Symbol comparisons: "List my closed XAUUSD trades, then compare their average net result with the overall period."
  • Loss reviews: "From the returned closed trades, identify the five largest net losses and calculate their average holding time."
  • Period comparisons: "Retrieve last month and this month separately, then compare trade count, win rate, and expectancy."

The server returns structured facts rather than trading signals. The connected model performs the explanation, so verify arithmetic and treat conclusions as review prompts—not instructions to enter or exit a market.

Zero-trust security: 100% read-only data isolation

Financial data privacy is critical. Edgelog was built from the ground up around strict data isolation principles, and the MCP integration follows a zero-trust model:

  • 100% read-only: The MCP server only answers inspection and aggregation queries. It contains zero trading endpoints and cannot place, alter, or cancel orders.
  • No credential exposure: Broker server addresses, account logins, investor passwords, exchange API keys, and personal journal reflections are strictly excluded from the tool interface.
  • Scoped to your account: Every incoming request must provide an authenticated bearer token. Queries are strictly bounded to the authenticated trader’s closed trades.
  • Hashed credentials: Plaintext tokens are displayed once upon creation and never stored. The database preserves only a one-way SHA-256 hash. You can revoke or rotate tokens with one click at any time.

Why native MCP beats manual CSV exports

Most traders who experiment with AI analysis export spreadsheet files and manually upload them to web chat interfaces. That workflow has severe limitations:

  1. Outdated context: The moment you close a new trade, your uploaded CSV is obsolete, forcing you to re-export and re-upload.
  1. Context window limits: Large CSV files quickly overwhelm language model context windows, resulting in truncation and hallucinated trade numbers.
  1. Security risks: Raw export files often contain broker account numbers, ticket IDs, and deposit balances that you may not want to share with generic AI web portals.

With Edgelog MCP, a compatible assistant queries specialized tools (get_trading_overview and list_closed_trades). Server-side date limits, symbol filters, and pagination reduce unnecessary disclosure and context usage. They do not guarantee that a language model will interpret the results correctly.

Supported AI assistants and environments

Remote MCP support and authentication options change across AI products and plans. Edgelog currently issues a read-only Bearer connection key, so a client must support a remote MCP endpoint with an authorization header. Do not paste the key into an ordinary chat message.

  • Claude products and developer tools: Anthropic documents MCP support across several Claude products, but remote-server configuration differs by product.
  • ChatGPT custom apps: OpenAI documents remote MCP custom apps for eligible plans and workspaces. Availability and accepted authentication mechanisms depend on the current product surface and administrator settings.
  • Developer clients: Custom agents can call the remote endpoint using the generated Bearer key and the MCP methods exposed by Edgelog.

A client should be listed as verified only after Edgelog has tested its current production release. Protocol support alone does not prove that a specific plan or interface accepts personal Bearer credentials.

Connect your trading journal to AI in two minutes

Getting started requires no technical expertise and no credit card. Log into your free Edgelog account, navigate to the workspace sidebar, and click Connect Edgelog MCP.

Click Generate key and copy the secret shown once. Add the endpoint and key through the client’s secure MCP settings—not inside a conversation. If the client requires OAuth instead of a Bearer header, it is not yet compatible with this access-key beta. Rotate or revoke the key from the Edgelog sidebar whenever access is no longer needed.

Frequently asked questions

QWhat is the Model Context Protocol (MCP)?

The Model Context Protocol is an open standard that allows AI assistants like Claude and Cursor to safely query structured external data sources without manual file uploads.

QCan an AI assistant place or modify trades on my broker account?

No. Edgelog MCP is 100% read-only. It only provides deterministic analytics and sanitized closed-trade histories. It has no access to broker execution.

QAre my broker passwords or exchange API keys shared with AI?

No. Broker logins, passwords, exchange keys, and private journal notes are strictly excluded from the MCP data projection.

QWhich AI assistants support Edgelog MCP?

Compatibility depends on whether the current client and plan support remote MCP servers with a Bearer authorization header. Edgelog should only describe a client as verified after testing its production configuration; generic MCP support alone is not enough.

QIs the Edgelog MCP integration free?

The current MCP beta is available without a separate subscription, subject to Edgelog account limits, request rate limits, and the 90-day lifetime of each connection key.

Log your trades. Master your edge.

Free for forex and crypto traders — up to 3 connected accounts and 50,000 stored trades, with MT4/MT5 auto-sync, exchange import, and full analytics.

Create your free journal