Probe is a fully local, AI-friendly semantic code search engine designed for large codebases, aiming to be a foundational tool for next-generation AI coding assistants. It targets developers and AI models needing precise, context-aware code retrieval.

How It Works

Probe combines the speed of ripgrep for initial file scanning with the structural understanding of tree-sitter for accurate code parsing. It then applies NLP techniques and re-rankers (BM25, TF-IDF, hybrid) to identify and extract complete code blocks like functions or classes, providing rich context for AI.

Quick Start & Requirements

• Install: npm install -g @buger/probe (also available via curl or PowerShell scripts).
• Requirements: macOS, Linux, or Windows; x86_64 or ARM64 architecture.
• AI Chat: npx -y @buger/probe-chat (requires API key for Anthropic/OpenAI).
• Docs: https://github.com/buger/probe

Highlighted Details

• Fully local and privacy-focused.
• Extracts entire functions/classes for AI context.
• Supports CLI, MCP server, and AI Chat modes.
• Integrates with AI editors via MCP.
• Offers hybrid, BM25, and TF-IDF re-ranking.

Maintenance & Community

• Active development with releases managed via GitHub Actions.
• Community support via Discord.

Licensing & Compatibility

• MIT License.
• Compatible with commercial and closed-source applications.

Limitations & Caveats

• AI Chat requires external API keys (Anthropic/OpenAI).
• Tree-sitter support is language-dependent; new languages require grammar additions.