Reference
Local semantic search for AI agents
Reference is on-device semantic search over your files and code, built for AI coding agents. It keeps a live index that updates as you save, uses code-aware chunking via tree-sitter, and ships an MCP server exposing search, explain, find_similar, and check_doc_drift. Results are cited down to the exact function rather than paraphrased, and nothing leaves the machine.
What is Reference?
Reference is on-device semantic search over files and code, built for AI coding agents, with a live index and an MCP server for cited, function-level results.
Key features
- Live index that updates as files are saved
- Code-aware chunking via tree-sitter
- MCP server exposing search, explain, find_similar, and check_doc_drift
- Results cited down to the exact function
- Fully local — nothing leaves the machine
- macOS app built with Tauri
Who it's for
- AI coding agents that need cited, targeted code context instead of repeated grep loops
- Teams whose docs drift from the code and need an automated staleness check
- Developers on large repos where token spend on context retrieval is a real cost
When not to use it
Not much benefit on a codebase small enough that a grep already fits comfortably in the agent's context window.
FAQ
What is check_doc_drift?
An MCP tool that flags when documentation has fallen out of sync with the code it describes.
Does Reference send code to a server?
No — indexing and search run entirely on-device.
Share this launch
Embed this badge
<a href="https://orangebot.ai/product/reference" target="_blank" rel="noopener noreferrer"> <img src="https://orangebot.ai/api/badge/reference.svg" alt="Featured on OrangeBot" width="200" height="54" /> </a>