Question 1

Do I need a GPU or special hardware to run vecgrep?

Accepted Answer

No. vecgrep uses Ollama with nomic-embed-text by default, which runs efficiently on CPU. Embedding generation is a one-time cost per indexing run — search itself is pure vector similarity and is instant on any machine.

Question 2

How is vecgrep different from grep or ripgrep?

Accepted Answer

grep and ripgrep find exact text patterns. vecgrep finds semantically related code — you describe what you are looking for in natural language and get results that match the meaning, not just the text. The hybrid mode also blends BM25 keyword matching so you never lose exact-match capability.

Question 3

Does my source code get sent to the cloud?

Accepted Answer

No. With the default Ollama provider, embeddings are generated locally and vectors are stored under ~/.vecgrep/projects/. Nothing leaves your machine. Cloud providers (OpenAI, Cohere, Voyage AI) are optional and only activated when you explicitly configure them.

Question 4

Which AI assistants work with the MCP server?

Accepted Answer

Any client that supports the Model Context Protocol, including Claude Code, Cursor, and custom MCP-compatible clients. The server exposes 11 tools for searching, indexing, status inspection, similar-code discovery, and codebase overview.

Question 5

What languages does vecgrep support?

Accepted Answer

vecgrep chunker is language-aware and supports Go, JavaScript, TypeScript, Python, Rust, Java, C/C++, and more. Code is chunked by syntax structures (functions, classes, methods) rather than arbitrary line splits.

Question 6

How does incremental indexing work?

Accepted Answer

vecgrep hashes every file on each vecgrep index run. Files whose hash has not changed are skipped — only new or modified files get re-embedded. A full rebuild (vecgrep index --full) is only needed when you change embedding model, dimensions, or chunking profile.

Question 7

Is vecgrep free and open source?

Accepted Answer

Yes. vecgrep is MIT-licensed and available on GitHub. Contributions, issues, and feature requests are welcome.

Studio (removed) ​

Studio (removed)