trace-mcp

trace-mcp is an MCP server (a connector that lets AI apps use another tool) that indexes a codebase once so coding agents stop re-reading files, for developers.

What it is good for

  • Reducing token costs when an AI agent reviews a pull request
  • Finding what breaks when changing a model or function
  • Tracing a web request through routes, services, and views
  • Recalling past decisions and sessions when starting a new coding task

At a glance

Who it is for
Setup difficulty
Technical ยท Guided command-line setup

Level 4 of 5. You copy and paste commands into a terminal by following the instructions. Comfortable territory for vibe coders.

Cost
Not stated
Type
MCP server or connector
Tasks
Works with
Claude Code, Cursor, Codex, Windsurf, Zed, Claude Desktop

How to get started with trace-mcp

  1. Install it globally with npm install -g trace-mcp.
  2. Run trace init once per machine to wire it into your agent.
  3. Run trace add inside the repo you want indexed.
  4. Ask your agent questions about the code, such as what breaks if you change a model.

From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.

Before you try it

  • Requires npm, so Node.js must be installed.
  • The README mentions a Max tier, so some features may be paid.
  • Benchmark results come from one build tested on six open-source repos.
  • Review quality scored close to plain file loading, with slightly more false positives per pull request.

Momentum

Rank this week
#68
Best rank
#68 week 46, 2026
Stars gained last week
+203
GitHub stars
204

GitHub numbers as of Nov 16, 2026. Updated weekly.

0 110 220 Sep 24, 2026 Nov 16, 2026
GitHub stars for trace-mcp: 179 on Sep 24, 2026, 204 on Nov 16, 2026. Source: GitHub.

Ranking history

First ranked in week 46, 2026.

Questions

Do I need to know how to code to use trace-mcp?
You do not need to write code, but setup means copying and pasting commands into a terminal by following its instructions.
What does trace-mcp work with?
According to its own documentation, it works with Claude Code, Cursor, Codex, Windsurf, Zed, and Claude Desktop.
How do I get started with trace-mcp?
1. Install it globally with npm install -g trace-mcp. 2. Run trace init once per machine to wire it into your agent. 3. Run trace add inside the repo you want indexed. 4. Ask your agent questions about the code, such as what breaks if you change a model.
What should I know before trying trace-mcp?
Requires npm, so Node.js must be installed. The README mentions a Max tier, so some features may be paid. Benchmark results come from one build tested on six open-source repos. Review quality scored close to plain file loading, with slightly more false positives per pull request.
What is trace-mcp good for?
Reducing token costs when an AI agent reviews a pull request, finding what breaks when changing a model or function, tracing a web request through routes, services, and views, and recalling past decisions and sessions when starting a new coding task.

Similar agents to trace-mcp

  • Repowise

    Also a codebase index for agents, but described without framework-specific integrations or decision memory.

  • xerj

    A local search engine and MCP server focused on finding code, docs and memory rather than building a framework-aware graph.

  • graphify

    A skill that builds a queryable knowledge graph of code and docs instead of running as an MCP server.

  • Understand-Anything

    A plugin that shows an interactive graph to help newcomers explore a project, rather than serving agents.

  • Grounded Docs MCP Server

    Serves searchable library documentation matching your versions, not an index of your own codebase.

  • AOCI-CODE

    AOCI-CODE is an MCP server (a connector that lets AI apps use another tool) that keeps a Git-versioned map of a codebase so coding agents stop re-reading it.

Also compared with trace-mcp on these pages:

Last updated Nov 16, 2026. Tracked since Oct 29, 2026. Parts of this summary required judgment because the source material was brief. Spotted an error? Tell us.