Codemogger
Code indexing library and MCP server that uses tree-sitter for semantic chunking and local embeddings, storing everything in a single SQLite file
318
GitHub Stars
13
Languages
1 File
SQLite DB
Overview
Codemogger is a code indexing library and MCP server designed specifically for AI coding agents. It parses source code using tree-sitter, chunks it into semantic units (functions, classes, impl blocks), embeds them locally using the all-MiniLM-L6-v2 model, and stores everything in a single SQLite file with both vector and full-text search capabilities. The tool requires no Docker, no external server, and no API keys—just one .db file per codebase. Codemogger returns the 5 most relevant definitions instead of thousands of matches, making it ideal for providing precise context to AI agents.
The Verdict
Who Should Use Codemogger?
Best For
- Developers building AI coding agents that need codebase context
- Teams wanting a simple, portable code index (single SQLite file)
- Users who need both keyword and semantic search combined
- Projects requiring incremental indexing (only re-embed changed files)
- Anyone integrating code intelligence via MCP protocol
Not Ideal For
- Enterprise teams needing cross-repository search at scale
- Projects requiring advanced code navigation (go-to-definition)
- Languages not supported by tree-sitter grammars
- Teams needing a managed, hosted solution
What's Great
- Zero dependencies—no Docker, no server, no API keys needed
- Single SQLite file stores everything (portable and simple)
- Tree-sitter parsing extracts semantic units, not arbitrary chunks
- Combines vector search and full-text search in one database
- Incremental indexing only processes changed files (SHA-256 hash)
- MCP server exposes search, index, and reindex tools
Watch Out For
- Newer project with smaller community
- Limited to 13 languages with tree-sitter grammars
- Large items (150+ lines) are subdivided, which may affect some searches
- No cloud or team collaboration features
Pricing
View all features & details
Key Features
- Tree-sitter semantic parsing
- Local embeddings (all-MiniLM-L6-v2)
- SQLite vector + FTS search
- Incremental indexing
- MCP server integration
- .gitignore-aware scanning
Supported Languages
- Rust, C, C++, Go, Zig
- Python, Java, Scala
- JavaScript, TypeScript, TSX
- PHP, Ruby
MCP Tools
- codemogger_search - semantic/keyword search
- codemogger_index - index a directory
- codemogger_reindex - force reindex
Installation
- npm install -g codemogger
- npx -y codemogger (no install)
- Library, CLI, or MCP server
How It Compares
| Feature | Codemogger | GrepAI | Sourcegraph |
|---|---|---|---|
| Parser | Tree-sitter (semantic) | Custom + embeddings | SCIP |
| Storage | SQLite (single file) | Custom index | Database |
| Search | Vector + FTS | Semantic | AI + Deterministic |
| MCP Server | Yes | Yes | Yes |
| Best For | Simple, portable indexing | Privacy-first search | Enterprise scale |
User Reviews
Loading reviews...