VectorCode
Code indexing for LLMs: VectorCode builds repository search and feeds relevant context to an agent via CLI and MCP
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Reads and indexes your repository files
- Sets up a local vector database and embedding model
- An external embedding engine, if connected, sends data out
Install
Manual install
uv tool install "vectorcode[mcp]<1.0.0"Install with MCP support via uv. Then index the repository with the vectorcode command.
This is third-party code. Review the repository files before installing.
What it does
VectorCode indexes a code repository and returns fragments relevant to the current task on demand. This helps build better context for coding LLMs, especially for closed-source and lesser-known projects the model knows little about, and reduces hallucination. It chunks files with syntax awareness via tree-sitter, respects .gitignore and detects the project root by anchors like .git or .vectorcode. It offers a command-line tool, a neovim plugin and an MCP server that wires the context into agents. Storage is built on ChromaDB.
Who it is for. For developers who want to give coding agents accurate context from their repository.
Good fit when
- You want to inject relevant pieces of your code into the prompt
- The agent knows a closed-source or niche project poorly
- You need local code search via RAG without manual copying
Not a fit when
- The project is small enough to fit whole in context
- You cannot set up a local vector database and embedding model
- You need a stable release; the project is beta and changes fast
Example request
Find the code that handles authorization in the indexed repository and add it to contextLimitations
The project is beta and iterates fast, interfaces may change, and the authors suggest pinning a version with a constraint. It needs Python 3.11 or newer, and building ChromaDB components may require recent c++ and rust compilers. By default embeddings run on the local SentenceTransformer model; other engines and acceleration need separate setup. The MCP mode installs an extra dependency group.
How to disable. Remove the server entry from your MCP client config. The tool itself is removed with uv tool uninstall vectorcode.
MCP
- Transport
- stdio
- Authentication
- not required
Security check
- Reads and indexes your repository files
- Sets up a local vector database and embedding model
- An external embedding engine, if connected, sends data out
README in short
The README presents VectorCode as a code repository indexing tool that helps build context for coding LLMs and reduces hallucination on lesser-known projects. It ships a command-line tool, a neovim plugin with an API set and an MCP server. Install is recommended via uv tool install, with extra dependency groups for LSP and MCP modes. It outlines a roadmap: syntax-aware chunking, .gitignore handling, project root anchors and collection deletion. The project is beta, storage is ChromaDB, and it is MIT licensed.
FAQ
Is it a CLI or an MCP server?
Both. There is a command-line tool for indexing and queries, a neovim plugin and an MCP server that feeds context to agents.
Does data leave your machine?
By default embeddings run on a local model and the index sits in a local ChromaDB database. An external embedding engine is wired in separately.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything