Vector Code MCP Server

VectorCode indexes source repositories so coding assistants can receive context relevant to the project and task at hand. The README frames this around code that models may not already understand well, such as private repositories, less familiar projects, or recent changes. Retrieved project material can be injected into prompts programmatically instead of expecting a model to infer details from its training alone. The repository contains both a command-line tool and a Neovim plugin. The plugin offers APIs for building or extending AI integrations, alongside integrations with existing editor plugins. MCP interfacing is acknowledged, although the supplied README does not enumerate MCP tool names or show a server launch command.

Completed features listed in the README include file metadata, configurable chunk sizes, syntax-aware chunking using tree-sitter, respect for gitignore rules, and project-root detection through .vectorcode or .git anchors. Collection management includes viewing and deleting indexed files, while asynchronous caching and persistent-client support are also marked implemented. Authenticated remote ChromaDB support and query rewriting remain unfinished items. The project is explicitly beta and changes quickly. Its pre-1.0 versioning scheme uses minor releases for breaking changes and patch releases for compatible updates or deprecation warnings. Setup instructions are linked in separate CLI and Neovim documentation, and users are told to consult documentation matching their installed version.