
stacklit
by glincker·★ 81·Score 46
Stacklit provides an MCP server that scans codebases to generate compact indexes for AI coding assistants.
Overview
Stacklit is a tool that transforms large codebases into compact, structured indexes (~250 tokens) that replace 50,000+ tokens of exploration. It parses code with tree-sitter for 11+ languages to extract module structure, dependencies, exports, and type definitions. The generated stacklit.json file provides AI agents with immediate codebase context without requiring file-by-file scanning. It includes MCP server functionality exposing 7 tools for retrieving overview, modules, dependencies, and hints.
Try asking AI
After installing, here are 5 things you can ask your AI assistant:
When to choose this
Choose Stacklit when working with AI coding assistants on large codebases where context window efficiency is critical and you need structured code navigation without overwhelming the agent with raw code.
When NOT to choose this
Don't choose Stacklit if you're working with very small codebases where the overhead of maintaining an index outweighs the benefits, or if you need real-time code analysis without any pre-processing.
Comparable tools
Installation
Install Stacklit
npm install -g stacklit # install globally
stacklit init # scan codebase
stacklit serve # start MCP serverConfigure Claude Desktop with MCP:
{
"mcpServers": {
"stacklit": {
"command": "stacklit",
"args": ["serve"]
}
}
}Configure Claude Code:
stacklit setup claude # updates CLAUDE.md + .mcp.jsonFAQ
- Does Stacklit read my code?
- Yes, locally only. It parses source files with tree-sitter to extract structure (imports, exports, types). No code is sent anywhere unless you use the optional --summary flag.
- What if my language isn't supported?
- Stacklit falls back to basic support (line count + language detection) for any language not in the tree-sitter list. The module map, dependency graph, and git activity still work.
Compare stacklit with
Last updated · Auto-generated from public README + GitHub signals.