SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI agentscapmonster-mcp-captcha-solver

capmonster-mcp-captcha-solver

Identify CAPTCHA types and parameters, create solving tasks and retrieve results via the CapMonster Cloud API, with access to its docs.

Open website

About this tool

Purpose

CapMonster Cloud MCP Server is a server based on the Model Context Protocol (MCP), used to let AI agents solve CAPTCHAs on their own. As the official MCP implementation of CapMonster Cloud, it provides a Python package as well as a TypeScript port (located in the ts/ directory), and can be used with Claude Desktop, Cursor, and other MCP clients.

Main capabilities

The server provides the following features:

  • List the CAPTCHA types supported by CapMonster Cloud.
  • Provide agents with CapMonster's live documentation.
  • Create solving tasks and poll task results against the CapMonster Cloud REST API.

Supported CAPTCHA types include (subject to the authoritative list returned by the get_supported_tasks tool from the live OpenAPI specification):

  • reCAPTCHA (v2, v2 Enterprise, v3)
  • Cloudflare Turnstile and Cloudflare Challenge (managed challenge / cf_clearance)
  • FunCaptcha (Arkose)
  • GeeTest (v3 and v4)
  • Enterprise-grade anti-bot systems: AWS WAF, DataDome, Imperva, TSPD, Binance, Prosopo, Yidun, TenDI, Hunt, Altcha, Basilisk, etc.
  • Image-to-text and complex image tasks

Note: hCaptcha is currently not supported.

Usage

The server is only responsible for the solving step and has no browser itself. It needs to be paired with an MCP that drives a browser (for example capmonster-mcp-patchright) to complete the page work (navigation, interaction, reading the live DOM/network, and injecting the solved token back into the page). For the complete step-by-step process, refer to capmonster_agent/SKILL.md.

Usage steps:

  1. Obtain an API Key.
  2. Add the server to the MCP client configuration.
  3. Ask the agent to open the CAPTCHA-protected page and solve it.

Runrequirements: Python 3.11+ and a valid CapMonster API Key. It can be run with uvx (no local clone required). If you prefer TypeScript/Node, see the npm-published port in ts/README.md—same tools, same behavior.

When used with an MCP client (such as Claude Desktop), add the configuration to the MCP client's configuration file (for example claude_desktop_config.json). The server runs only over stdio, so there are no HTTP headers to carry a per-request key—CM_API_KEY is read once from the environment variable and shared across every tool call in the session.

It is recommended to pair it with a browser automation MCP server (for example capmonster-mcp-patchright), so that the agent can both view the page and solve the CAPTCHA on the page—see the example in mcp.example.json that configures both servers together.

Available MCP tools

Once connected, the LLM can access the following tools:

  • get_supported_tasks: List the CAPTCHA task types supported by CapMonster from the live OpenAPI specification.
  • get_task_parameters(task_type): Get the required/optional fields, variant descriptions, and solution schema for a task type.
  • get_docs(url, offset, limit, section): Get CapMonster documentation pages (limited to docs.capmonster.cloud / api.capmonster.cloud), with support for section jumps and pagination.
  • create_task(task): Submit a CAPTCHA task and return a taskId.
  • get_task_result(task_id): Poll once for a task result.
  • get_task_result_wait(task_id, timeout_seconds, poll_interval_seconds): Poll a task until it completes (better than driving your own polling loop).
  • get_actual_user_agent(): Get the current Windows User-Agent, used for both the browser and the solving taskconsistent fingerprint between tasks.
  • get_balance(): Check the current CapMonster Cloud API balance.

Use Cases

Suitable for automation workflows that require AI agents to automatically handle CAPTCHAs, such as testing, data collection, or automated operations on CAPTCHA-protected pages (must be used on websites for which you have permission).

Limitations

  • This server itself does not include a browser and must be used together with a browser automation MCP.
  • It runs only via stdio; the API Key is read once from an environment variable and shared throughout the session.
  • hCaptcha is currently not supported.
  • Requires Python 3.11+ and a valid CapMonster API Key.
  • Should only be used on websites you own or have permission to automate, and in compliance with the relevant website's terms of service and applicable laws.

Official Documentation

  • CapMonster Cloud MCP page
  • CapMonster Cloud main documentation
  • Model Context Protocol (MCP) documentation

Related tools

Browse category