MCP · Cloudflare edge

Web search for agents.
At the speed of thought.

Faaast is a search, scrape and grounded-research server for AI agents. One MCP endpoint, 13 sharp tools, edge-cached results, and answers stripped of filler — so your model spends tokens on the work, not the warm-up.

Get a free key Read the docs No credit card. 2,000 calls/day.

Connect in one line

Streamable HTTP MCP with OAuth 2.1 + PKCE. Claude Code, Cursor, VS Code and the MCP Inspector discover the login flow automatically — no config, no proxy.

claude mcp add --transport http faaast \
  https://faaast.ereneksi.com/mcp

Or paste a header

Prefer API keys? Create one in the dashboard and send it as a bearer token. Keys are hashed at rest and revocable in one click.

{
  "mcpServers": {
    "faaast": {
      "url": "https://faaast.ereneksi.com/mcp",
      "headers": { "Authorization": "Bearer faa_sk_…" }
    }
  }
}

13 tools, one contract

Every tool returns compact text plus structured JSON. Every result carries usage counters. Nothing returns a wall of caveats.

web_search

Search the live web.

research

One call: search, read the best sources in parallel, and return only the passages that answer the question, with citations.

fetch_url

Fetch one page and return clean, readable markdown (boilerplate, nav, scripts and ads removed).

batch_fetch

Fetch up to 8 URLs in parallel and return clean markdown for each.

news

Recent news and announcements, ranked by recency and relevance.

extract

Pull exact fields out of a page with CSS selectors (title, price, author, links, table cells) or JSON-LD.

sitemap

Discover a site's URLs from its robots.txt and XML sitemaps (handles sitemap indexes).

wikipedia

Authoritative encyclopedia summaries from Wikipedia's API.

github_search

Search GitHub repositories, issues/PRs, code and users via the official API.

hackernews

Search Hacker News stories and discussion via the Algolia API.

arxiv

Search arXiv preprints by title, abstract or author.

package_info

Exact metadata for an npm or PyPI package: latest version, license, description, repo, dependency count, release date.

account_usage

Current fair-use counters for this account: calls used today, per-minute headroom, reset time.

Accurate by construction

Results are fused across independent search backends with reciprocal-rank fusion, then scored on term coverage, phrase match, recency and source authority. Content farms are demoted; primary sources win.

research goes further: it searches, reads the top sources in parallel, and returns only the passages that answer the question — with citations already attached.

Fast where it matters

Edge-cached search~3 ms
Live search (fused)200–600 ms
Page read + markdownnetwork-bound
HTML → markdownnative, streaming
Auth checkedge-cached

Plays by the rules

Faaast reads robots.txt before every fetch and honours Content-Signal: ai-input=no and X-Robots-Tag: noai. It identifies itself honestly, never bypasses a paywall or login, never solves a CAPTCHA, and caches aggressively so origins see less traffic.

Search uses official APIs — Brave, Tavily, Exa, or key-free public APIs (Wikipedia, arXiv, GitHub, Hacker News, Stack Exchange). No search-engine HTML scraping.

Fair use, stated plainly

Free2,000 calls/day · 60/min
Pro40,000 calls/day · 400/min
account_usagefree
Cached responsesstill 1 call (metered)

Limits are enforced per account in a Durable Object. No surprise throttling, no silent degradation — you get an exact reset time.

Point your agent at it

Free account, instant key, OAuth or bearer.

Create account API reference