Hey PH! I built Index to solve a friction point I kept hitting with AI coding assistants. When you're working with libraries that have sparse documentation, recent API changes, or just weren't well-represented in training data—AI assistants struggle. They'll suggest methods that don't exist in your version, or miss the idiomatic patterns entirely. The fix: let the AI search your actual dependency source before writing code. Index parses your dependencies (npm, pip, cargo), builds a local semantic index, and exposes it via MCP. Now when Claude Code needs to use a library, it searches the index first and gets real function signatures from your exact installed version. Works especially well for newer libraries, niche packages, or anything with fast-moving APIs. Curious what libraries give your AI tools the most trouble—would love to hear.

Index

Semantic search pinned to your dependency versions

Semantic search pinned to your dependency versions