codebase-memory-mcp is a high-performance code intelligence MCP (Model Context Protocol) server that indexes codebases into a persistent knowledge graph. It provides sub-millisecond query responses and supports 158 programming languages with zero dependencies, delivered as a single static binary.

Installation is simple with our automated installers:

For more installation options and manual installation instructions, visit our Installation Guide.

codebase-memory-mcp auto-detects and configures 11 popular coding agents:

• Claude Code
• Codex CLI
• Gemini CLI
• Zed
• OpenCode
• Antigravity
• Aider
• KiloCode
• VS Code
• OpenClaw
• Kiro

The install script automatically configures MCP entries, instruction files, and pre-tool hooks for each detected agent.

Indexing is extremely fast:

• Linux kernel (28M LOC, 75K files) in 3 minutes
• Django project (49K nodes) in ~6 seconds
• RAM-first pipeline with LZ4 compression and in-memory SQLite
• Memory released after indexing completes

Our benchmarks show we index 120x faster than traditional file-by-file approaches.

We support 158 languages through vendored tree-sitter grammars. Key languages include:

• Python, TypeScript/JavaScript, Go, Rust, Java, C++, C, PHP, C#
• Enhanced with Hybrid LSP semantic type resolution for 9 languages
• Infrastructure-as-code: Docker, Kubernetes, Kustomize
• Benchmarked against 64 real repositories with excellent coverage

Full language list available in our Features documentation.

14 powerful MCP tools for code analysis:

• Indexing: index_repository, list_projects, delete_project, index_status
• Querying: search_graph, trace_path, detect_changes, query_graph, get_graph_schema
• Analysis: get_architecture, search_code, manage_adr, get_code_snippet

Each tool provides structured data for AI agents to understand codebases efficiently.

Hybrid LSP combines tree-sitter's syntactic parsing with lightweight semantic type resolution algorithms inspired by major language servers (tsserver, pyright, gopls, rust-analyzer, etc.). It provides:

• Import-aware resolution across modules
• Type inference and generic handling
• Inheritance and trait method resolution
• JSX component dispatch and JSDoc inference

Available for Python, TypeScript/JS, PHP, C#, Go, C/C++, Java, Kotlin, and Rust.

Our knowledge graph represents code as interconnected nodes and edges:

• Nodes: Project, Package, File, Function, Class, Method, Interface, Route, etc.
• Edges: CALLS, IMPORTS, DEFINES, IMPLEMENTS, HTTP_CALLS, DATA_FLOWS
• Persistence: SQLite database with WAL mode for ACID safety
• Query: Cypher-like queries with relationship traversal

This structure enables ultra-fast structural analysis and impact mapping.

Several optimizations enable large codebase handling:

• RAM-first pipeline with LZ4 compression
• Incremental indexing with background watcher
• Project artifact sharing (.codebase-memory/graph.db.zst)
• Parallel indexing with configurable worker count
• Efficient in-memory SQLite with single dump at end

We've successfully indexed the Linux kernel (28M LOC) in just 3 minutes.

Yes! codebase-memory-mcp processes 100% locally:

• No external API calls or network dependencies
• All processing happens on your machine
• Knowledge graph stored locally in ~/.cache/codebase-memory-mcp/
• Embeddings bundled in binary (nomic-embed-code)

Your code never leaves your machine - ideal for security-sensitive environments.

We implement multiple security layers:

• Binary Signing: All releases are ad-hoc signed on macOS
• CodeQL SAST: Static analysis blocks releases with vulnerabilities
• VirusTotal Scanning: 70+ antivirus engines scan every release
• SLSA Level 3: Cryptographic build provenance
• Supply Chain: Zero runtime dependencies, all libraries vendored

See our Security Policy for details.

No. codebase-memory-mcp processes everything locally:

• All indexing and analysis happens on your machine
• No data is sent to external servers
• No API keys required
• No telemetry or analytics
• Knowledge graph stored locally only

Your codebase privacy is our priority - we never transmit or store your code externally.

Check these common issues:

• Ensure the binary is in your PATH or use absolute paths
• Restart your agent after installation
• Verify MCP configuration file paths
• Test binary: echo '{}' | /path/to/codebase-memory-mcp

For specific agent issues, check our Multi-Agent Support Guide.

Use the built-in update command:

The server also checks for updates on startup and notifies you when a new release is available.

To remove configurations:

This removes all agent configs, skills, hooks, and instructions but keeps the binary and databases. To remove everything: