← All guides
Agent integrations

Auxiliar MCP: Web Access for Any Agent (Quickstart)

Give any MCP client — Claude Code, Cursor, or your own agent — web search and scraping through the Auxiliar gateway on one API key.

Updated 2026-06-30 · Auxiliar

The Model Context Protocol (MCP) is how agents discover and call tools. If your agent speaks MCP — Claude Code, Cursor, Windsurf, or your own client — the fastest way to give it the whole web is to connect it to the Auxiliar gateway: one key that reaches every search, scraping, crawling and voice provider.

Option 1 — the Auxiliar MCP

Add the Auxiliar MCP to an MCP-capable client:

claude mcp add auxiliar -- npx auxiliar-mcp
# or, in any client's MCP config:
#   command: npx   args: ["auxiliar-mcp"]

The MCP connects your agent to the Auxiliar gateway and tells it how to reach every provider on your key. The full in-loop tool surface — web_search, scrape, extract, crawl executed over the gateway and billed to your credits — is rolling out toward v1.0; the docs always reflect the current tool set.

Option 2 — the gateway as HTTP tools (works today, everywhere)

Any agent that can make an HTTP request can use Auxiliar right now — no MCP required. Expose two tools that call the gateway:

import os, requests
AUX = "https://api.auxiliar.ai"
H = {"Authorization": f"Bearer {os.environ['AUXILIAR_API_KEY']}"}

def web_search(query: str) -> dict:
    """Search the live web."""
    return requests.post(f"{AUX}/serper/search", headers=H, json={"q": query}).json()

def scrape(url: str) -> dict:
    """Fetch a page as clean markdown."""
    return requests.post(f"{AUX}/firecrawl/v1/scrape", headers=H,
                        json={"url": url, "formats": ["markdown"]}).json()

Register those with your framework’s tool interface (or wrap them in a tiny MCP server) and the agent has search + scrape on one credential.

Why the gateway is the right MCP backend

  • One key, every provider. No agent should manage 24 provider accounts. Auxiliar injects the upstream keys server-side and bills one balance.
  • Route to the best per job. Search with Tavily or Serper, scrape with Firecrawl or Scrapfly — each a benchmark winner — by changing a path.
  • Fallback on failure. When a scraper is blocked, the agent retries through another provider on the same key. See scrape without getting blocked.

Get a key at accounts.auxiliar.ai, then point your MCP client — or any agent — at the gateway.

One key. Every provider on this page.

Stop juggling signups and invoices. One Auxiliar API key calls all of them — upstream keys injected server-side, usage billed to a single balance. Swap the base URL and go.

curl https://api.auxiliar.ai/serper/search \
  -H "Authorization: Bearer $AUXILIAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"q": "latest ai agent news"}'

Keep building