Solve CAPTCHAs in CrewAI with MCP (No Composio Required)

CrewAI has native MCP support. GateSolve has an MCP server. Put them together and your crew can solve CAPTCHAs without a single line of custom integration code.

The Problem

Your CrewAI agent needs to scrape a site, fill a form, or access data behind a Cloudflare Turnstile challenge. The agent hits the CAPTCHA and stops. You need a CAPTCHA solver, but wiring up API calls inside CrewAI tools is tedious and brittle.

2Captcha solved this by getting listed on Composio, so CrewAI users can install it as a connected MCP. But Composio adds a dependency layer. CrewAI already supports MCP servers natively — you can skip the middleware.

The Solution: GateSolve MCP + CrewAI

CrewAI's mcps field on agents accepts MCP servers directly. GateSolve's MCP server (@gatesolve/mcp-server) exposes solve_captcha and detect_captcha as tools. Connect them and the LLM handles the rest.

Option 1: Structured Config (Recommended)

from crewai import Agent, Task, Crew
from crewai.mcp import MCPServerStdio

# Create an agent with GateSolve MCP tools
scraper = Agent(
    role="Web Scraper",
    goal="Extract data from CAPTCHA-protected pages",
    backstory="Expert scraper that handles CAPTCHAs automatically",
    mcps=[
        MCPServerStdio(
            command="npx",
            args=["-y", "@gatesolve/mcp-server"],
            env={"GATESOLVE_API_KEY": "gs_your_key_here"}
        )
    ]
)

# The agent now has solve_captcha and detect_captcha tools
task = Task(
    description="""Go to https://example.com/protected-page.
    If there is a CAPTCHA, detect its type and solve it.
    Then extract the main content.""",
    agent=scraper,
    expected_output="Extracted page content"
)

crew = Crew(agents=[scraper], tasks=[task])
result = crew.kickoff()
print(result)

Option 2: String Reference (Quick Setup)

If you prefer the DSL approach, you can reference the MCP server directly — though you'll need the server running locally or via a remote endpoint:

scraper = Agent(
    role="Web Scraper",
    goal="Scrape CAPTCHA-protected sites",
    backstory="Handles CAPTCHAs via GateSolve",
    mcps=[
        MCPServerStdio(
            command="npx",
            args=["-y", "@gatesolve/mcp-server"],
            env={"GATESOLVE_API_KEY": "gs_your_key_here"}
        )
    ]
)

What Tools Does the Agent Get?

When you connect @gatesolve/mcp-server, the agent automatically discovers two tools:

• solve_captcha — Submit a CAPTCHA for solving. Pass the URL, CAPTCHA type (turnstile, recaptcha, hcaptcha), and siteKey. Returns the solution token.
• detect_captcha — Detect what type of CAPTCHA a URL has before attempting to solve. Returns the type and siteKey.

The LLM decides when to call each tool based on the task description. No hardcoded logic needed.

Multi-Agent Crews

CrewAI shines with multi-agent workflows. You can give the CAPTCHA solver to one specialist agent and let others focus on their tasks:

from crewai import Agent, Task, Crew
from crewai.mcp import MCPServerStdio

gatesolve_mcp = MCPServerStdio(
    command="npx",
    args=["-y", "@gatesolve/mcp-server"],
    env={"GATESOLVE_API_KEY": "gs_your_key_here"}
)

# Agent 1: Handles CAPTCHAs
unblocker = Agent(
    role="CAPTCHA Unblocker",
    goal="Detect and solve CAPTCHAs on target URLs",
    backstory="Specialist in bypassing web protections",
    mcps=[gatesolve_mcp]
)

# Agent 2: Extracts data (no CAPTCHA tools needed)
extractor = Agent(
    role="Data Extractor",
    goal="Extract structured data from web pages",
    backstory="Expert at parsing HTML and extracting information"
)

# Agent 3: Analyzes results
analyst = Agent(
    role="Data Analyst",
    goal="Analyze extracted data and produce insights",
    backstory="Expert data analyst"
)

# Chain tasks
solve_task = Task(
    description="Check https://example.com for CAPTCHAs and solve them",
    agent=unblocker,
    expected_output="CAPTCHA solution token or confirmation page is accessible"
)

extract_task = Task(
    description="Extract pricing data from the unblocked page",
    agent=extractor,
    expected_output="Structured pricing data"
)

analyze_task = Task(
    description="Analyze pricing trends from extracted data",
    agent=analyst,
    expected_output="Pricing analysis report"
)

crew = Crew(
    agents=[unblocker, extractor, analyst],
    tasks=[solve_task, extract_task, analyze_task]
)
result = crew.kickoff()

Why MCP Over Manual API Calls?

Get Started

1. Get a free API key at gatesolve.dev — 100 free solves, no credit card
2. Install CrewAI: pip install crewai crewai-tools
3. Add the MCP server to your agent's mcps config
4. Run your crew — the agent handles CAPTCHAs automatically

Supported CAPTCHA types: Cloudflare Turnstile, reCAPTCHA v2/v3, and hCaptcha. Average solve time: ~12 seconds.

Compare: GateSolve vs 2Captcha via Composio

• 2Captcha + Composio: Install Composio SDK → connect 2Captcha → import toolkit → wrap as CrewAI tools. Three dependencies, Composio account required.
• GateSolve MCP: Add MCPServerStdio to agent mcps list. One dependency (npx), no middleware account.

Both work. GateSolve is a shorter path with fewer moving parts. Plus 100 free solves to test before committing.