For developers

Your Claude agents, with a browser of their own.

Add UnblockingAPI's hosted MCP server to a Messages API request and Claude can read any page and pull JSON from templates while it answers - no MCP client to run, one entry in mcp_servers.
Connect Claude API

500 free credits  ·  1 credit per successful fetch  ·  Open-source MCP server

python
import os
import anthropic

client = anthropic.Anthropic()

response = client.beta.messages.create(
    model="claude-opus-5-5",
    max_tokens=2000,
    messages=[{"role": "user", "content": "Get the Google results for 'espresso machine' and list the top 5."}],
    mcp_servers=[{
        "type": "url",
        "url": "https://mcp.unblockingapi.com/mcp",
        "name": "unblockingapi",
        "authorization_token": os.environ["UNBLOCKINGAPI_KEY"],
    }],
    tools=[{"type": "mcp_toolset", "mcp_server_name": "unblockingapi"}],
    betas=["mcp-client-2025-11-20"],
)
print(response.content[-1].text)

No client to run

Anthropic's API connects to the server for you; you pass the URL and a token.

Pages and data

Claude fetches rendered, unblocked pages and gets JSON where a template covers the site.

Your account, your limits

Calls run on your UnblockingAPI key, with your plan's credits and concurrency.

The Claude API's MCP connector (beta mcp-client-2025-11-20) lets a Messages request use tools from a remote MCP server. Point it at https://mcp.unblockingapi.com/mcp with your UnblockingAPI key as the authorization token and Claude gets unblock_fetch, find_templates and get_build_guide.

Call it from the Messages API

One mcp_servers entry

RecommendedURL + API key

Same hosted URL, authenticated by your key in a header.

curl https://api.anthropic.com/v1/messages \
  -H "content-type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "anthropic-beta: mcp-client-2025-11-20" \
  -d '{
    "model": "claude-opus-5-5",
    "max_tokens": 2000,
    "messages": [{"role": "user", "content": "Read https://example.com and summarise it."}],
    "mcp_servers": [{
      "type": "url",
      "url": "https://mcp.unblockingapi.com/mcp",
      "name": "unblockingapi",
      "authorization_token": "'"$UNBLOCKINGAPI_KEY"'"
    }],
    "tools": [{"type": "mcp_toolset", "mcp_server_name": "unblockingapi"}]
  }'
The Python SDK version is on the right. Parameter names follow Anthropic's MCP connector docs.

Try it in Claude API

Use this as the user message in your request. Claude should find the Google template and return parsed results.
Use unblockingapi to get the Google results for "best espresso machine" as structured data, then summarise the top three.
No code

Want structured data instead of raw HTML?

unblock_fetch hands back the full rendered page. For clean fields instead - a price, a title, a whole list - build a template visually in the editor: load the page, tap what you want, done. No CSS selectors to write by hand.

Save it and it's a reusable API endpoint - same account, same key, same as the built-in templates. Build the structure once, then call it from your app - or your AI agent - forever.

Try the editor

No account needed to try it  ·  No credit card  ·  More about the editor

amazon.comSample data

Select one item. Similar items are detected automatically.

Website

Sony WH-1000XM5

4.6

$299.99

Bose QuietComfort

4.5

$249.00

Bang & Olufsen Beoplay

4

$1,714.99

Extracted JSON

{ "products": [
  { "name": "Sony WH-1000XM5", "price": 299.99, "currency": "USD", "rating": 4.6 },
  { "name": "Bose QuietComfort", "price": 249, "currency": "USD", "rating": 4.5 },
  { "name": "Bang & Olufsen Beoplay", "price": 1714.99, "currency": "USD", "rating": 4 }
] }

Tap one item, get every item - as an array.

Claude API questions

Your UnblockingAPI API key. Anthropic sends it to our server as a bearer token on every call, so tool calls run on your account. Keep it server-side, like your Anthropic key.

Yes - the mcp_toolset entry takes an allowlist or denylist of tool names. See Anthropic's MCP connector docs.

The MCP connector when Claude decides what to fetch; the REST API or an SDK when your own code does. Both spend the same credits.

With the key header, the key is stored in Claude API's config, so keep that file out of source control. Where sign-in is available it is the safer option: the app then holds only a revocable token, and you can disconnect it under Connected apps.

The MCP server is free and open source. Fetches run on your UnblockingAPI account: every successful fetch costs 1 credit and failed ones are free. New accounts get 500 free credits with no credit card, and plans make credits cheaper at volume. Looking up templates is free.

Three, all read-only. unblock_fetch reads any URL through rotating residential proxies and a real browser, picking a plain fetch or a browser per site by itself. It returns the page as Markdown (HTML on request), or JSON with a template. find_templates checks whether a published template covers a site. get_build_guide gives Claude API the API reference and recipes when you build an app on the API. Details on the MCP server page.

The hosted server: nothing to install or update, and most apps sign in without a key. Run it locally with npx -y @unblockingapi/mcp if you prefer to keep the process on your machine - the tools and costs are the same.

Also works with

One account, every assistant. Connect as many as you like, or see them all.

n8n

MCP Client Tool node

Claude

Connector, plugin or one command

Cursor

One click, then sign in

VS Code

code --add-mcp, then sign in

Windsurf

mcp_config.json, then sign in

Zed

settings.json

Give your Claude agents a real browser.

500 free credits, no credit card. Connect in under a minute.