trace-mcp

Overview

trace-mcp is an MCP server designed to reduce AI agent recomputation by building a framework-aware dependency graph of codebases and serving it through the Model Context Protocol. It helps AI coding agents avoid re-reading the same files, re-traversing dependencies, and re-inflating context windows with previously discovered structure. The server offers significant benefits including 40-50% fewer tokens, up to 2× effective capacity, and up to 99% less redundant processing. It supports 81 framework integrations across 80 languages with 162 tools, and also indexes markdown knowledge vaults alongside code.

Try asking AI

After installing, here are 8 things you can ask your AI assistant:

• get_change_impact

  Returns blast radius across codebase when changing a specific symbol

  Try:What breaks if I change this model?Try:What's the impact of renaming this function?
• query_decisions

  Searches the decision knowledge graph linked to code

  Try:Why was auth implemented this way?Try:What decisions were made about the database schema?
• plan_turn

  Provides optimal code subgraph + relevant past decisions for a new task

  Try:I'm starting a new task, what should I know?Try:Help me plan the implementation of this feature
• search_sessions

  Searches across past session content for relevant information

  Try:What did we discuss about GraphQL last month?Try:Find previous conversations about authentication
• get_request_flow

  Returns request flow from URL to rendered page with framework awareness

  Try:Show me the request flow from URL to rendered pageTry:How does this API endpoint connect to the database?
• get_untested_symbols

  Identifies symbols that are untested or imported but never called in tests

  Try:Find all untested code in this moduleTry:Which methods don't have test coverage?
• get_subproject_impact

  Returns impact of API changes across different services

  Try:What's the impact of this API change on other services?Try:How will this change affect dependent microservices?
• find_usages

  Finds backlinks across codebase or markdown vault

  Try:What notes link to this concept?Try:Where is this function used in the codebase?
• apply_rename

  Renames symbols and updates all references across codebase

  Try:Rename this class and update all referencesTry:Change this function name everywhere it's used
• get_task_context

  Provides optimal code subgraph for understanding a feature

  Try:Give me the context for the user authentication featureTry:Show me the relevant code for the payment processing
• get_wake_up

  Provides orientation in ~300 tokens for new sessions

  Try:Give me a summary of what we were working onTry:What's the current state of this project?
• get_session_resume

  Carries over structural context from previous sessions

  Try:Resume our previous work on this featureTry:What were we working on before?

Comparable tools

Installation

npm install -g trace-mcp

{
  "mcpServers": {
    "trace": {
      "command": "trace-mcp",
      "args": []
    }
  }
}

FAQ

What frameworks and languages does trace-mcp support?

trace-mcp supports 81 framework integrations across 80 languages, including PHP, Vue, Laravel, React, Angular, and many more. It understands framework-specific patterns and constructs dependency graphs accordingly.

Can trace-mcp work with documentation and markdown files?

Yes, trace-mcp can index markdown knowledge vaults like Obsidian and Logseq. It treats wikilinks, tags, frontmatter, and embeds as graph edges and metadata, allowing the same tools to work across both code and documentation.

How does trace-mcp improve AI agent performance?

trace-mcp reduces AI agent recomputation by providing pre-computed dependency structures instead of requiring agents to repeatedly read files and traverse dependencies. This leads to fewer tokens, lower latency, higher accuracy, and better scaling with project size.

Compare trace-mcp with