Episkey-G/GrokSearch-rs

What it solves

grok-search-rs is a lightweight Rust-based Model Context Protocol (MCP) server that enables AI assistants to perform live web searches and fetch page content. It bridges the gap between LLMs (specifically Grok/xAI and OpenAI-compatible gateways) and the live web, providing a structured way to integrate search and retrieval tools into an agent's workflow.

How it works

The server exposes a set of tools (web_search, get_sources, web_fetch, web_map, doctor) that an MCP client can call. It supports two transport modes: local stdio (where the client launches the server) and remote HTTP (for multi-device access).

It utilizes a pluggable source chain to gather information, attempting to use providers in a specific order (Tavily, Exa, TinyFish, Firecrawl) to ensure reliability. For specific sites like GitHub, StackExchange, arXiv, and Wikipedia, it uses specialist extractors to produce clean Markdown instead of generic scraping.

Who it’s for

Developers and users of MCP-compatible AI assistants (like Claude Code) who want to give their agents the ability to search the web, extract clean content from technical documentation and forums, and integrate with Grok or other OpenAI-compatible LLM gateways.

Highlights

• Dual Transport Support: Runs locally via stdio or as a remote multi-tenant HTTP service.
• Response Budgeting: Manages token limits by trimming responses and paging through cached sources.
• Specialist Extractors: High-quality Markdown parsing for GitHub, StackExchange, arXiv, and Wikipedia.
• Pluggable Source Chain: Ordered fallback mechanism across multiple search providers (Tavily, Exa, TinyFish, Firecrawl).
• Flexible Auth: Supports both API keys and xAI OAuth mode for native Grok access.