XERJ

XERJ is an open-source local search engine from xerj-org on GitHub, and it doubles as an MCP server you can connect to coding agents such as Claude Code and Cursor. It's a single Rust binary that indexes code, docs, logs, JSONL, CSV and PDFs for full-text, vector and hybrid search, and it can jump straight to the file and line where a function is defined. The author says its main use is "reference coding": before writing code, download the open-source repositories closest to the task, index them and see how others solved the same problem. The README cites a controlled study in which this cut a coding agent's output tokens by 2.7 times at the same solve rate. What sets it apart is that it gets the agent to look things up before it writes.

Who it's for:developers on large projects who want their coding agent to read fewer files and spend fewer tokens. Readers who don't write code won't have a use for it.

Before you start: the README's quick start downloads an install script with curl and pipes it straight to sh, which runs a remote script directly on your computer, so we don't recommend it. Follow the "Install by hand" section instead. The README also offers a one-prompt install for AI agents that has the agent download tools and clone repositories on its own; we don't recommend that either, so do the steps yourself. The --insecure flag in the example commands turns off TLS and API-key checks, so anyone who can reach the port can read and delete every index. The author says a fresh install listens only on your own machine; don't open it to the internet. The README also asks users and AI agents to send the project a short field report as a pull request, and your agent may be asked to do this even when it's working in someone else's codebase. Whether to send one, and what goes in it, should be your call, not something the agent does on its own. Reference coding clones other people's repositories, so check their licences before reusing code. Apache-2.0 licence.

Based on the project's public documentation. Not tested by AI Inspiration World.