unch, created by Uchebnick, is a local-first semantic code search tool that helps developers locate functions, methods, and variables inside large repositories. It runs natural-language and keyword searches against a locally built index and can serve as a Model Context Protocol server so AI assistants can query a project's code. Key capabilities include vector and keyword search modes, multi-language parsing, and CI support for shared indices. The tool targets engineers and DevOps professionals who need fast, private, terminal-based code exploration.

What tasks can you actually use it for?

The tool supports terminal-driven code discovery, letting developers ask natural-language queries to find definitions, references, and inline documentation across a repository. It also functions as an MCP server so AI assistants can access local source for contextual prompts, which helps when generating or debugging code with a model that can query the index. Example tasks include locating a function implementation, finding usages of a method, and surfacing nearby documentation for quick context.

How reliable are the search results for real codebases?

Search behavior changes by mode: the semantic mode uses vector matching to find conceptually related symbols, while the lexical mode returns exact text matches. Once an index exists, searches are near-instantaneous because they operate locally rather than over the network. Results depend on repository structure and supported languages; accuracy benefits when files are well-typed and use recognizable symbol names, and it performs best on the languages the tool optimizes for.

What inputs and constraints affect what it finds?

The indexer parses source files and documentation into searchable entries, so the contents and language grammars influence coverage. Supported languages include Go, Rust, TypeScript, JavaScript, and Python; additional languages require available grammars. The tool requires a terminal environment and a Node.js runtime for installation, and MCP client compatibility if used as a server. Large repositories require an initial index build before fast queries are available.

Is it practical to add to developer workflows and CI?

unch provides integration points for automation, including GitHub Actions to keep shared indices updated across teams, and distribution via npm or Homebrew for easy deployment. The cross-platform Node.js approach supports macOS, Linux, and Windows environments, and the command-line interface fits into shell-based workflows and scripts. Teams can automate index updates in CI to keep search results current for collaborators and AI tools.

A sensible choice for terminal-oriented teams that need local code visibility

For developers comfortable in a terminal environment who prioritize keeping source code on their machines and enabling AI assistants to query local context, this tool offers a practical, output-focused solution. Expect an up-front indexing step and some setup in CI for teamwide use. The tool suits projects where fast, private code exploration and model-assisted context are more important than a graphical search experience.