Logo
webctx

webctx

A high-performance Model Context Protocol server that supercharges AI agents with intelligent web search and content extraction. Aggregates results from Brave, Tavily, and Exa with intelligent ranking algorithms, extracts clean markdown from web pages, and includes an observability dashboard for monitoring tool usage.

Visit Project
Next.js 15 MCP Brave Search Tavily Firecrawl Turso

Features

  • Multi-provider search aggregation querying Brave, Tavily, and Exa in parallel
  • Score-based ranking algorithm weighting results by position, provider, and cross-provider duplicates
  • Intelligent content extraction with fast paths for GitHub and .md files
  • Site mapping to discover all indexed URLs on any website
  • Domain filtering with automatic exclusion of video/social sites
  • Observability dashboard with real-time tool call monitoring
  • API key authentication for production deployments
  • Queue-based rate limiting preventing API throttling

Challenge

AI agents need high-quality, up-to-date web information but single search providers often return inconsistent results. Building a reliable research tool requires aggregating multiple sources and intelligently ranking results.

Solution

Created an MCP server that runs parallel searches across Brave, Tavily, and Exa, then applies a sophisticated scoring system rewarding results appearing in multiple sources. Added fast paths for developer-focused content and comprehensive observability for debugging.

Highlights

  • Multi-provider aggregation delivers higher quality results
  • Cross-provider scoring surfaces most relevant content
  • Fast paths optimize for developer documentation