VecGrep FAQs

Name: VecGrep
Author: VecGrep

Question 1

What is VecGrep and how does it save me tokens?

Accepted Answer

VecGrep is a Claude Code plugin that performs semantic code search. Instead of sending entire files, it returns only the most semantically relevant code chunks to Claude, achieving up to a 95% token reduction for codebase queries. This saves costs and accelerates AI-powered code analysis.

Question 2

What embedding providers are supported, and can I use my own?

Accepted Answer

VecGrep offers flexible embedding options. It defaults to a local, efficient model (`all-MiniLM-L6-v2-code-search-512`) via fastembed. You can also configure it to use cloud providers like OpenAI, Voyage AI, or Google Gemini by providing your own API key.

Question 3

How does VecGrep intelligently find relevant code snippets?

Accepted Answer

It parses source files using tree-sitter to extract semantic units (functions, classes, methods), encodes these into embeddings with your chosen provider, stores them in LanceDB, and then performs a fast approximate nearest neighbor (ANN) search to find the most relevant chunks based on your query.

Question 4

How do I integrate VecGrep with Claude Code?

Accepted Answer

VecGrep integrates seamlessly as an MCP (Multi-Codebase Plugin) for Claude Code. After a simple installation (`claude mcp add --scope user vecgrep -- vecgrep`), Claude automatically calls VecGrep's tools (like `index_codebase` and `search_code`) based on your prompts, eliminating manual tool invocation.

Question 5

What programming languages does VecGrep support for semantic chunking?

Accepted Answer

VecGrep provides native semantic chunking for a wide range of languages including Python, JavaScript/TypeScript, Rust, Go, Java, C/C++, Ruby, Swift, Kotlin, and C#. Other text files are handled by intelligent sliding-window line chunking.

VecGrep

VecGrep

主要功能

使用案例

主要功能

使用案例