APIStatusGitHub
Cookbook

Anthropic via MCP

Call Krawl's MCP server from a Claude agent loop — drop Krawl into Anthropic's Messages API as a remote MCP server and Claude gains research, quick_search, and ask_krawl tools.

Anthropic's Messages API speaks MCP (mcp-client-2025-04-04 beta). Add Krawl as a remote server and any Claude conversation gains research-grade tools that return cited reports.

Working file: docs/cookbook/anthropic-via-mcp.py.

Install

pip install anthropic
export ANTHROPIC_API_KEY=sk-ant-...
export KRAWL_API_KEY=krawl_demo_xxx     # optional — lets Claude bypass 402

curl

You don't normally talk to MCP with curl, but you can hit the Krawl MCP endpoint to confirm it's alive:

curl -X POST https://api.krawl.sh/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

The reply lists every tool with its declared price (USDC, atomic units) and JSON Schema input.

Python

import os, sys
import anthropic

client = anthropic.Anthropic(api_key=os.environ["ANTHROPIC_API_KEY"])

response = client.beta.messages.create(
    model="claude-opus-4-7",
    max_tokens=2048,
    system="You are a research analyst. Use krawl.research for deep questions.",
    messages=[{"role": "user", "content": " ".join(sys.argv[1:])}],
    mcp_servers=[
        {
            "type": "url",
            "url": "https://api.krawl.sh/mcp",
            "name": "krawl",
            "tool_configuration": {
                "enabled": True,
                "allowed_tools": ["research", "quick_search", "ask_krawl"],
            },
            "authorization_token": os.environ.get("KRAWL_API_KEY"),
        }
    ],
    betas=["mcp-client-2025-04-04"],
)

for block in response.content:
    if hasattr(block, "text"):
        print(block.text)

Run:

python cookbook/anthropic-via-mcp.py "Compare EigenLayer and Symbiotic"

Node

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY! });

const res = await client.beta.messages.create({
  model: "claude-opus-4-7",
  max_tokens: 2048,
  system: "Use krawl.research for deep questions. Cite every claim.",
  messages: [{ role: "user", content: "Compare EigenLayer and Symbiotic" }],
  mcp_servers: [
    {
      type: "url",
      url: "https://api.krawl.sh/mcp",
      name: "krawl",
      tool_configuration: {
        enabled: true,
        allowed_tools: ["research", "quick_search", "ask_krawl"],
      },
      authorization_token: process.env.KRAWL_API_KEY,
    },
  ],
  betas: ["mcp-client-2025-04-04"],
});

for (const block of res.content) {
  if ("text" in block) process.stdout.write(block.text);
}

Notes

  • The MCP server exposes research, quick_search, and ask_krawl today. The full v1 surface adds list_lookouts, get_lookout, run_lookout, export_result, get_usage, get_memory, and compare_results — see Agents & MCP.
  • Skip authorization_token and Claude will pay per call via x402 — Krawl returns a 402 challenge inline. See Payments & x402.
  • For Claude Code (the CLI), install the skill directly: npx skills@1.7.4 add ravidsrk/krawl --skill krawl --agent claude-code (requires repository read access while private).

OpenAI Agents SDK

Run an o3-deep-research-shaped OpenAI Agents SDK agent with Krawl as the model — point the OpenAI client at /v1 and the response is a cited research report.

Vercel AI SDK — streamText

Stream a Krawl deep-research report through Vercel AI SDK 6 streamText — wire Krawl in via the OpenAI-compatible provider and pipe the cited markdown into a server component or CLI.

On this page

InstallcurlPythonNodeNotes