Claude, Cursor and MCP
Connect TraceX to Claude Code, Cursor or any MCP client to ask about production, start investigations and find the change behind an incident from your editor.
If your team already codes with an AI assistant such as Claude Code or Cursor, you don’t need to switch tools to use TraceX. TraceX is a remote MCP server: add it once, and your assistant can ask TraceX about production, start investigations and find the change behind an incident without leaving your editor.
Why connect your AI tool
Section titled “Why connect your AI tool”Your coding assistant can read your local files, but it can’t see production. It doesn’t know what’s in the database, what the logs say or what shipped yesterday. With TraceX connected, it can check before it answers or changes code:
- “Is
orders.coupon_idever null in production?” → it asks TraceX instead of guessing. - “Why is checkout failing?” → it starts an investigation and reads the result.
- “Did my last deploy cause this error?” → it asks for suspect changes.
TraceX acts as you, with your access, and reads only.
Connect
Section titled “Connect”-
In the dashboard, open Profile → AI tools and click Create a token. Copy it now: it is shown once. It works for a limited time and is listed under Profile → Sessions, where you can revoke it.
-
Add TraceX to your tool:
Terminal window claude mcp add --transport http tracex https://api.tracexpro.com/mcp/ \--header "Authorization: Bearer <your token>"In
~/.cursor/mcp.json:{"mcpServers": {"tracex": {"url": "https://api.tracexpro.com/mcp/","headers": { "Authorization": "Bearer <your token>" }}}}Streamable HTTP transport:
URL: https://api.tracexpro.com/mcp/Header: Authorization: Bearer <your token>
| Tool | What it does |
|---|---|
list_projects |
The projects you can use. Pass a project’s name to the other tools. |
ask_tracex |
Ask about production: data, code, logs, errors and deploys. |
start_investigation |
Start a full root-cause investigation; returns its id and link. |
get_investigation |
An investigation’s status and, when finished, its root cause and fix. |
suspect_changes |
Which deploys, commits and pull requests most likely caused an incident. |
similar_investigations |
Past investigations of similar issues, with their root causes. |
Example prompts
Section titled “Example prompts”Once connected, just ask your assistant in plain language. It picks the right tool:
- “Use TraceX to check how many users signed up yesterday by plan.”
- “Ask TraceX why order 48213 failed, then fix the bug you find.”
- “Start a TraceX investigation: EU uploads fail since 14:30.”
- “Which commits most likely caused the 09:02 checkout errors?”
- “Have we seen this error before? Check TraceX for similar investigations.”
Rules and limits
Section titled “Rules and limits”The same rules as Conversations apply: questions are checked by TraceX’s input scanner, skills are read-only, plan limits count, and answers are checked before they are returned. Questions appear in Conversations with an AI tool badge.
Revoke access
Section titled “Revoke access”Tokens work for a limited time. Revoke one any time under Profile → Sessions, and
remove TraceX from your tool’s MCP settings (for Claude Code: claude mcp remove tracex).