Federated search
Query Brave, Tavily, and Exa together, remove noisy domains, deduplicate URLs, and return a ranked markdown list.
Pure Go web context CLI
Give agents clean web search, markdown page reads, and site maps from one terminal command.
webctx is built for agent workflows that need reliable web context as text. It combines Brave, Tavily, and Exa search, reads links into markdown, and maps websites into URL lists that are easy for agents to inspect and reuse.
Install it with npm or build the native Go binary directly. Add provider keys through environment variables, `.env.local`, or macOS Keychain, then pipe the output wherever your agent works.
What it solves
Web context that is already shaped for agents.
Browser pages are noisy and provider APIs disagree. webctx normalizes common web-research tasks into markdown output: ranked links, extracted documents, and URL maps.
Markdown page extraction
Read public GitHub files through raw-content fast paths, fetch direct markdown pages, and fall back to Firecrawl when needed.
Site mapping
Use Firecrawl's map endpoint to discover URLs across a docs site or product site before choosing pages to read.
Agent-ready output
Every command prints plain text or markdown that can be pasted into coding agents, research notes, issue threads, or scripts.
Built for
Coding agents, research agents, and maintainers who prefer terminal-native context.
Search the web from a terminal agent without hand-curating browser tabs.
Turn GitHub README and blob URLs into markdown quickly without a scraping provider.
Fallback to Firecrawl for normal pages, PDFs, and site maps when raw markdown is not available.
Ship as a Go binary through GitHub Releases while keeping npm install convenient for CLI users.
Command model
Three commands, four providers, one clean output style.
search
Runs multi-provider search, filters domains, deduplicates normalized URLs, and ranks results.
read-link
Prefers GitHub raw content and direct markdown before using Firecrawl scrape fallback.
map-site
Calls Firecrawl map with sitemap inclusion, subdomains, query stripping, and a 5,000 URL limit.
distribution
Ships native Go binaries through GitHub Releases and installs through an npm postinstall shim.
Use webctx when an agent needs web evidence as text: search candidates, read the useful links, and map the site before deciding what to inspect next.