Vectorcode, developed by Davidyz for the mcp platform, is a specialized code repository indexing tool. Built primarily with Python, it connects to a vector database, ChromaDB, to store embedding vectors of codebases. It provides an experimental implementation of the Model Context Protocol (MCP) server, standardizing how applications deliver context to Large Language Models (LLMs). This allows developers to enhance LLM interactions by injecting relevant code context directly into prompts, improving accuracy and reducing model hallucinations.

What does Vectorcode connect to?

Vectorcode functions as a code repository indexing tool, primarily connecting to ChromaDB, a vector database. It stores embedding vectors derived from codebases within ChromaDB. Additionally, Vectorcode integrates with the Model Context Protocol (MCP) as an MCP server. This server implementation standardizes how various applications provide contextual information to Large Language Models. This setup allows for efficient storage and retrieval of code-related embeddings, serving as a critical backend for LLM context provision.

Core features and tools

Vectorcode's MCP server exposes several callable tools for managing code context. These include ls to list local collections, query for retrieving information from a specified collection, and vectorise to convert files into project-specific embedding vectors. Users can also employ files_ls to view indexed files for the current project and files_rm to remove selected files from the project database. These tools empower developers to interact directly with the indexed codebase.

Who is it for?

Vectorcode is primarily designed for developers working with Large Language Models, especially those using Neovim, who seek to enhance their LLM experience. It targets users aiming to provide repository-level context to LLMs, thereby improving model output quality. Compatible AI clients include mcphub.nvim, which works with codecompanion.nvim, and Avante, alongside Opencode. This tool is ideal for integrating structured code context into LLM workflows. Skip this if you do not work with LLMs or require repository-level code context.

How to get started

No credentials are required.

Use cases

• Improving LLM code completion and documentation by injecting task-relevant context from a codebase into prompts, thereby reducing hallucinations and increasing accuracy.
• Programmatically providing Large Language Models with relevant code context to significantly improve the quality and pertinence of model output.
• Assisting LLMs in debugging or exploring unfamiliar repositories by enabling them to actively search for relevant files and understand codebase structure.

Vectorcode offers a specialized solution for integrating codebase context with Large Language Models via the MCP server. Its ability to index code into a vector database and expose tools like query and vectorise makes it valuable for developers. This enhances LLM performance by providing direct, relevant information, reducing common issues like factual inaccuracies. It serves as a backend for sophisticated LLM interactions. Try querying an indexed collection to provide specific codebase context to your LLM.