MCP for AI agents

CapMonster Cloud MCP — Let AI Agents Solve CAPTCHA Automatically.

Connect CapMonster Cloud to an AI agent and automate CAPTCHA solving, from collecting parameters to checking the result. The agent can hand the working logic back to your project.

Start in a few minutes with a ready-made prompt.

Run a ready-made prompt in your AI agent

It will help you install and configure CapMonster Cloud MCP in your AI client. Open it in Claude Code or Codex, or copy it for another MCP-compatible client.

Codex MCP Server Setup

Open the setup prompt in Codex, then send it to start.

Other MCP-compatible clients

Choose an npm or PyPI prompt and paste it into your AI client. PyPI needs Python 3.11+ and uv. Browser automation still uses Node.js. If automatic setup is unavailable, use Manual setup.

HelpIcon
An AI agent may limit or reject MCP because of its own terms or rules. That does not mean MCP is broken. See the setup guide and the troubleshooting guide, or try another compatible client and follow that client's rules.

How to Connect CapMonster Cloud MCP

  1. Get an API key

    Create a CapMonster Cloud account and copy your API key.

  2. Choose how to connect

    Open the prompt in Claude Code or Codex, or copy it for another MCP client.

  3. Send the prompt to the agent

    The agent asks for the API key and stores it in CM_API_KEY.

  4. Check the installation

    Ask the agent to verify the setup, or run the /mcp command.

  5. Describe the task

    Give the page URL, the action that triggers CAPTCHA, and any extra access conditions.

Prefer to install MCP manually?

If automatic setup is not a fit, start the packages yourself. Configuration still depends on the client.

How CapMonster Cloud MCP works

From page to working integration — in one MCP session: the agent identifies the CAPTCHA, collects parameters, calls CapMonster Cloud, verifies the result, and prepares working code.

  1. AI agent
    1 . Identifies the CAPTCHA

    The AI agent opens the page through Patchright, finds the challenge, and determines the CAPTCHA type.

    Starting: opening the page through patchright and checking the CapMonster Cloud balance.

    Used 2 tools >

    Balance available, browser started. Opening the page with CAPTCHA.

    Patchright: navigate to page >

    Need to find the sitekey and reCAPTCHA parameters on the page.

  2. CAPTCHA
    2 . Gets the task parameters

    Analyzes the page and extracts the CAPTCHA type, sitekey, URL, and other widget data for CapMonster Cloud.

    Result

    CAPTCHA type and parameters:

    reCAPTCHA v2 (checkbox, "I'm not a robot"), protection level: high

    data-sitekey = 6Lcg7CHUAAAAANphynKgn9YAgA4tQ2KI_iqRyTwd

    websiteURL = https://lessons.zennolab.com/captchas/recaptcha/v2_simple.php?level=high

    Widget in iframe google.com/recaptcha/api2/anchor – standard reCAPTCHA v2 (not Enterprise or invisible)

  3. CapMonster
    3 . Creates a task and gets the result

    Sends the parameters to CapMonster Cloud, creates a task, and waits for the response. The agent receives the token automatically.

    Creating a task in CapMonster Cloud: RecaptchaV2Task, no proxy — the site is publicly accessible.

    Capmonster: create task >

    Task created (ID 643066485). Waiting for the solution for up to 1 min.

    Capmonster: wait for task result >

    Token received. Injecting into g-recaptcha-response , calling the callback and submitting the form.

  4. Working logic
    4 . Verifies the result and prepares the code

    Applies the token on the page, confirms the challenge passed, and saves a code sample for the project.

    Token injected. Checking the value and clicking "Submit".

    Patchright: run code on the page >

    Value set correctly. Submitting the form.

    CAPTCHA passed: the server returned "success": true . Closing the browser and preparing a code sample.

    </>solve_recaptcha_v2.pyUndo+125-0>

The agent works with CapMonster Cloud and the browser

Two MCP servers cover one workflow. capmonster talks to the API. patchright talks to the page.

capmonster

Work with CapMonster Cloud

  • Loads documentation
  • Suggests task parameters
  • Creates tasks
  • Returns the result
  • Checks the balance and User-Agent
patchright

Work with the page in the browser

  • Interacts with the page
  • Detects the CAPTCHA type
  • Gets dynamic parameters
  • Applies the result
  • Checks that the logic works

Example scenarios

How an AI agent helps with real CAPTCHA work.

Understand a new CAPTCHA

Show the agent a page — it will help identify the CAPTCHA type and the required parameters.

Check an integration

The agent will send the task, get the result, and check it right on the page.

Add solving to a project

After a successful check, the agent will help turn the working logic into code.

Find the cause of an error

It will check parameters, the User-Agent, cookies, headers, and other session data.

Which CAPTCHAs AI Agents Solve with CapMonster Cloud MCP

CapMonster Cloud supports many popular protection and verification systems.

Work with MCP and without it

The same steps — finding parameters, choosing a task, waiting for the result, and preparing the integration — go through MCP tools instead of manual work.

Without MCP
What the developer does manually

  • Finding parameters

    Studies DevTools, requests, and page code.

  • Choosing the task type

    Learns CAPTCHA types and studies the documentation.

  • Waiting for the result

    Implements result polling and response handling.

  • Integrating into the project

    Assembles working code for the site and environment.

Via MCP
What the AI agent does with CapMonster Cloud MCP

  • Getting parameters

    Gets the required parameters through the browser MCP and DevTools.

  • Choosing the task type

    Picks the task type from the get_supported_tasks list.

  • Getting the result

    get_task_result_wait automatically waits for the finished answer.

  • Preparing the integration

    After a successful check, prepares code for your project.

Advantages of CapMonster Cloud MCP

Fast setup, less manual work, and CAPTCHA automation in your projects and AI clients.

Less manual work

No need to look up parameters separately, check every step against the documentation, or write result polling just for the first check.

Less code for integration

Automate API integration: the agent calls CapMonster Cloud tools, gets the list of supported tasks, creates a task, and waits for the result.

Work in your familiar client

Use MCP in compatible AI clients without a separate integration for each interface.

One API key

The agent only needs CM_API_KEY to access CapMonster Cloud. When browser work is needed, it connects through a separate MCP server.

CapMonster Cloud Pricing

success icon
Proxy included in the price: save on CapMonster Cloud*
CAPTCHA
Price (USD)
reCAPTCHA v2-logoreCAPTCHA v2
tick99%
$ 0.04*
1000 images
$ 0.60
1000 tokens
reCAPTCHA v3-logoreCAPTCHA v3
tick99%
$ 0.90
1000 tokens
reCAPTCHA v2 Enterprise-logoreCAPTCHA v2 Enterprise
tick97%
$ 0.04*
1000 images
$ 1.00
1000 tokens
reCAPTCHA v3 Enterprise-logoreCAPTCHA v3 Enterprise
tick98%
$ 1.50
1000 tokens
Cloudflare Turnstile-logoCloudflare Turnstile
tick99%
$ 1.30
1000 tokens
Cloudflare Bot Challenge-logoCloudflare Bot Challenge
tick99%
$ 1.30
1000 tokens
Amazon Waf-logoAmazon Waf
tick99%
$ 1.40
1000 tokens
Tencent CAPTCHA-logoTencent CAPTCHA
tick99%
$ 1.60
1000 tokens
GeeTest-logoGeeTest
tick98%
$ 1.20
1000 tokens
Yidun-logoYidun
tick95%
$ 1.00
1000 tokens
Text CAPTCHA-logoText CAPTCHA
tick99%
$ 0.30
1000 tokens

Frequently asked questions about CapMonster Cloud MCP

CapMonster Cloud MCP connects an AI agent to CAPTCHA solving. The capmonster server works with the CapMonster Cloud API: documentation, task parameters, creating a task, and the result. The patchright server works with the browser: it opens the page, reads dynamic parameters, applies the solution, and checks that the page accepted it.

Choose a ready-made prompt for Claude Code, Codex, npm, or PyPI and send it to the agent. The agent asks for your API key and stores it in CM_API_KEY. You can also start the packages yourself with npx or uvx.

npm / TypeScript runs with npx -y capmonster-mcp. PyPI runs with uvx capmonster-mcp and needs Python 3.11+ and uv. Both expose the same solve tools. Browser automation uses the npm package capmonster-mcp-patchright and still needs Node.js.

Both servers are recommended for a full browser workflow. capmonster solves the task. patchright opens the page, collects parameters, and verifies the result. You can use capmonster on its own when the page parameters are already known.

Learn more – in the official CapMonster Cloud documentation.

Useful Materials

Guides, documentation, and product pages for getting started with CapMonster Cloud.

Product
Cloudflare Bot Challenge

How the challenge works and how CapMonster Cloud solves it.

Product
CAPTCHA Solver Browser Extension

Solve CAPTCHA in Chrome and Firefox without writing a client.

Product
reCAPTCHA v2

Parameters, task type, and a working integration path.

Documentation
CapMonster Cloud MCP Documentation

Reference for both MCP servers, prompts, and tools.

Guide
How to Set Up and Use CapMonster Cloud MCP

A walkthrough of npm, PyPI, Claude Code, and Codex.

Article
Why Do Websites Think I’m a Bot?

How detection systems use headers, fingerprints, and session data.

Developer Resources

More CapMonster Cloud Tools

CapMonster Cloud API

Direct access to the service for integration into your projects

Browser extensions

Solve CAPTCHA right in your browser

n8n integration

Ready-made nodes and examples for automation

Ready to try CapMonster Cloud MCP?