codebase-memory-mcp

Overview

codebase-memory-mcp is a cutting-edge code intelligence tool that transforms codebases into structured knowledge graphs through advanced tree-sitter AST analysis. It stands out for its extreme indexing speed - able to process the Linux kernel (28M LOC, 75K files) in just 3 minutes - and offers 14 MCP tools for comprehensive code exploration. The server provides zero-dependency operation through a single static binary across multiple platforms, with built-in 3D graph visualization UI available at localhost:9749. Its research-backed approach demonstrates significant efficiency improvements, reducing token usage by 10x and tool calls by 2.1x compared to traditional file-by-file exploration methods.

Try asking AI

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

• get_architecture

  Returns languages, packages, entry points, routes, hotspots, boundaries, layers, and clusters

  Try:What is the architecture of this project?Try:Show me the entry points and hotspots in the codebase
• manage_adr

  Persists architectural decisions across sessions

  Try:Create a new architectural decision recordTry:Show me all architectural decisions
• detect_changes

  Maps uncommitted changes to affected symbols with risk classification

  Try:Show me the impact of my uncommitted changesTry:What symbols will be affected by these changes?
• semantic_query

  Vector search across the entire graph using embedded code embeddings

  Try:Find functions related to payment processingTry:Search for code that handles user authentication
• search_graph

  Searches the graph with regex patterns, filters, and degree constraints

  Try:Find all functions with 'Handler' in the nameTry:Show me nodes with more than 10 connections
• search_code

  Graph-augmented grep over indexed files only

  Try:Find all occurrences of 'processOrder'Try:Search for functions that use 'lodash' utilities
• trace_call_path

  Resolves function calls across files and packages

  Try:What calls the ProcessOrder function?Try:Show me the call path from main to database connection
• detect_dead_code

  Finds functions with zero callers, excluding entry points

  Try:Show me unused functions in the codebaseTry:Find dead code that can be removed
• index_repository

  Indexes the current repository into the knowledge graph

  Try:Index this projectTry:Build the codebase memory
• get_cross_repo_links

  Shows links between nodes across multiple indexed repositories

  Try:Show dependencies between this service and othersTry:What other repositories reference our API?
• find_http_routes

  Finds HTTP routes and their implementations in the codebase

  Try:Show all REST endpoints in the applicationTry:Find route handlers for user management
• find_channels

  Detects pub-sub patterns across services and modules

  Try:Find event emitters and listeners in the codeTry:Show pub-sub patterns between microservices

Comparable tools

Installation

curl -fsSL https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.sh | bash

# 1. Download the installer
Invoke-WebRequest -Uri https://raw.githubusercontent.com/DeusData/codebase-memory-mcp/main/install.ps1 -OutFile install.ps1

# 2. (Optional but recommended) Inspect the script
notepad install.ps1

# 3. Run it
.\install.ps1

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

FAQ

What languages are supported?

codebase-memory-mcp supports 155 programming languages through vendored tree-sitter grammars compiled into the binary. It offers enhanced LSP-style type resolution for Go, C, C++, and TypeScript/JavaScript.

How does it handle large codebases?

It uses a RAM-first pipeline with LZ4 compression and in-memory SQLite, processing the Linux kernel (28M LOC) in 3 minutes. Memory is released after indexing, and it includes auto-sync for change detection.

Is it secure?

Yes - all processing happens 100% locally with no external API calls. Every release binary is signed, checksummed, and scanned by 70+ antivirus engines. Full source code is available for audit.

On Hacker News

Recent discussion from the developer community.

• I replaced grep-based code exploration with a knowledge graph – 10x less token▲ 4💬 3
  Story by DeusData · 2026-03-03

Compare codebase-memory-mcp with