Reflex
Local-first, full-text code search MCP server for AI coding agents. Indexes your codebase with trigram-based indexing for sub-100ms queries. Agents can search across 10k+ files, filter by language or symbol kind, and get structured JSON results — all offline, no API key required
Documentation
>>Reflex
Local-first code search engine with full-text search, symbol extraction, and dependency analysis for AI coding workflows.
$ npm install -g reflex-search
Why Reflex
Fast at Scale
Indexes a 27,448-file Kubernetes checkout in 7.7s on 16 cores. Median queries on a 30 MB corpus: ~3ms for a common word, ~12ms for a regex.
Complete Coverage
Finds every occurrence — not just definitions — across the same files ripgrep searches. Whole-identifier matching by default, plus substring and case-insensitive search.
AI-Native
MCP server with 17 tools, JSON output for agent pipelines, built-in AI query assistant. In A/B tests, same task success at ~31% lower cost per task.
15 Languages + Text
Rust, TypeScript, Python, Go, Java, C/C++, and more, parsed with Tree-sitter for precise symbols. Every other non-binary file is indexed for full-text search.
Dependency Analysis
Import tracking, reverse lookups, circular detection, hotspot analysis across your entire codebase.
Deterministic
Fully local, fully offline. Same query, same results, and every response says whether the index is fresh. Your code stays on your machine.
See it in action
Install
npm npm install -g reflex-search
cargo cargo install reflex-search
shell (Linux, macOS) curl --proto '=https' --tlsv1.2 -LsSf https://github.com/reflex-search/reflex/releases/latest/download/reflex-search-installer.sh | sh
PowerShell (Windows) powershell -ExecutionPolicy Bypass -c "irm https://github.com/reflex-search/reflex/releases/latest/download/reflex-search-installer.ps1 | iex"