Tools

MCP Tools

The server exposes 16 tools across 10 engines, allowing LLMs to choose the right tool for the specific task.

search_web: Performs universal web and news searches. Use this for quick, broad discovery. Supports domain-scoped search via the domain parameter (e.g., domain="docs.python.org").
fetch_web_page: Extracts clean, main-content text from a given URL. Supports multiple output formats (txt, json, markdown, html, xml). Ideal for reading articles.

search_exa: Semantic web search via Exa AI. Uses sentence-level matching instead of keywords. Supports content type filters (company, research paper, news, tweet), date ranges, and domain scoping. Optionally requires EXA_API_KEY for higher rate limits (free tier: 20K requests/month).

search_reddit: Search Reddit via keyless RSS and shreddit enrichment. Excellent for finding community sentiment, real-user experiences, and discussions. Supports subreddit targeting, depth control, and time filters.

search_hackernews: Search Hacker News via the Algolia API. Great for developer opinions, startup discussions, and technical news. No API key needed. Supports depth control for comment enrichment.

search_github: Search GitHub Issues and PRs across repositories. Great for bug reports, feature requests, and community sentiment on open-source projects. Optionally uses GITHUB_TOKEN or gh CLI for higher rate limits.
get_github_issue: Fetch a full GitHub Issue or PR thread with all comments sorted by reactions. Requires gh CLI installed and authenticated.

search_wikipedia: Search Wikipedia and return full article text with related results. Great for factual, encyclopedic context. No API key needed.

search_arxiv: Search arXiv for academic papers by keyword, author, or category. Supports Lucene field prefixes (au:, ti:, abs:, cat:) for targeted searches. No API key needed.

search_x: Search X/Twitter via vendored Bird CLI. Great for real-time discourse, breaking news, and community engagement signals. Requires AUTH_TOKEN and CT0 environment variables from browser cookies.

groq_search: AI-powered web search. Two modes: GPT-OSS models for interactive browsing, Compound models for auto-research. Best for deep research and multi-step synthesis.
groq_analyze: Visits a URL and interprets its content in a single step. Uses Groq Compound to fetch and analyze in one call.

get_package_info: Look up a specific package from npm, PyPI, crates.io, or Go modules. Returns version, description, downloads, license, dependencies count, and repository info.
search_packages: Search packages by keyword across npm, PyPI, crates.io, or Go. Returns ranked list with metadata.
analyze_error: Parse an error message with auto-detected language/framework, then search Stack Overflow for solutions. Extracts file paths and line numbers from stack traces.
compare_technologies: Side-by-side technology comparison using GitHub stars, registry versions, download counts, license, and open issues.

Task	Tool
Quick web search	`search_web`
Semantic / deep search	`search_exa`
Read an article	`fetch_web_page`
Community opinions	`search_reddit` or `search_hackernews`
Code/issue lookup	`search_github`
Full issue/PR thread	`get_github_issue`
Academic papers	`search_arxiv`
Encyclopedia / facts	`search_wikipedia`
Real-time / breaking news	`search_x`
Package info / search	`get_package_info` / `search_packages`
Debug errors	`analyze_error`
Compare technologies	`compare_technologies`
Deep research	`groq_search`
URL analysis	`groq_analyze`