2Captcha MCP

What 2Captcha MCP does

2Captcha MCP MCP server exposes web-access tools to AI agents through the Model Context Protocol. Its capabilities include ranked web search results, page retrieval as Markdown or HTML, marketplace data extraction, custom JSON-schema extraction, background batch jobs, CAPTCHA solving, and selected browser workflows.

The server belongs in the 🔎 Search & Data Extraction category. It is suited to agents that need current information from sites that may restrict ordinary HTTP clients. Marketplace parsing can return fields such as product title, price, rating, seller, stock, and offers. Structured extraction accepts a URL or text together with a user-provided JSON Schema.

CAPTCHA support covers the types supported by 2Captcha, including reCAPTCHA, Turnstile, hCaptcha, DataDome, and image challenges. The browser-related tools can identify a CAPTCHA from HTML supplied by a browser session and return JavaScript or a cookie for use in that same session. An opt-in login workflow can sign in through a managed browser and preserve the resulting session.

How it works

The 2Captcha MCP MCP server can run remotely at https://mcp.2captcha.com/mcp or locally through npx @2captcha/mcp. Remote connections use an HTTP Authorization bearer header. Local deployments communicate over stdio and receive the credential through the API_TOKEN environment variable.

For page scraping, requests move through escalating retrieval methods. The documented sequence can try hidden JSON APIs, a TLS-impersonating HTTP client, rotating residential proxies, and a managed browser with CAPTCHA solving. Marketplace extraction similarly tries deterministic sources such as marketplace APIs, JSON-LD, and learned selectors before using an LLM extraction step when needed.

Long-running collection can be submitted as a batch job, then monitored or cancelled. The account tool reports remaining allowance. The hosted and local deployments expose the same general tool surface, although the connection method differs.

Setup and configuration

Create or obtain a 2Captcha API key from the account settings page. For a local client configuration, use the official scoped package @2captcha/mcp, set API_TOKEN, and launch it with npx. Do not substitute the unrelated unscoped 2captcha-mcp npm package.

A remote MCP client should connect to the hosted URL and send Authorization: Bearer YOUR_API_TOKEN. A local configuration follows this pattern:

{
  "command": "npx",
  "args": ["@2captcha/mcp"],
  "env": { "API_TOKEN": "YOUR_API_TOKEN" }
}

The README documents setup examples for Claude Desktop, Claude Code, Cursor, Codex CLI, and VS Code. Claude Desktop can also install a release .mcpb extension. A free account includes a rolling allowance of 200 calls or $0.50 in measured spend per 30 days, whichever is reached first; CAPTCHA solves remain charged to the 2Captcha balance.

Tools and capabilities

• Search the web and return ranked organic results with titles, URLs, and snippets.
• Scrape URLs as clean Markdown or raw HTML.
• Parse marketplace pages and listings into structured product data.
• Extract JSON that conforms to a supplied schema.
• Submit, poll, and cancel multi-URL background jobs.
• Detect and solve supported CAPTCHA types, including challenges in an existing browser session.
• Log in through an opt-in managed browser workflow and save the session.
• Inspect account allowance and report call metadata when requested.

Limitations and notes

The 2Captcha MCP MCP server requires an account credential; it does not provide anonymous access. Usage is subject to the account allowance and 2Captcha balance. Browser-session CAPTCHA solving returns material to apply in the caller's own session rather than moving that session into the managed browser. The README does not establish support for every MCP client, so clients not listed should be tested with the standard stdio or HTTP connection mode they provide.