[ AGENTS ]

Captcha solving for
MCP agents

An agent that meets a challenge has nothing in its toolset that can answer it, so it stalls or hallucinates its way around. Give it a solver as a tool and the block becomes one more call.

json · mcp config

{
  "mcpServers": {
    "captchano": {
      "command": "npx",
      "args": ["-y", "@captchano/mcp"],
      "env": { "CAPTCHANO_API_KEY": "cn-YOUR_API_KEY" }
    }
  }
}

// Exposes one tool: solve_captcha(type, sitekey, url, proxy?)
// Returns { token | cookie, userAgent, cost, solveTime }

Why agents need this specifically

A scraper that hits a challenge fails loudly and you fix it. An agent fails quietly: it reasons about the blocked page, invents a plausible next step, and reports progress it never made. The failure is worse because it is silent. A typed solver tool turns an ambiguous dead end into a deterministic result the agent can branch on.

What the tool returns

A token or cookie — depending on the challenge type. Session-graded systems return a cookie plus the header set used to earn it.
The user agent used, so the agent can replay it on submission.
The cost of the solve, so an agent operating under a budget can account for it rather than spending blind.

Keeping the loop bounded

Give the agent a solve budget per run. Every response reports its cost, so a supervisor can cut the loop when spend passes a threshold rather than discovering it on the invoice. At $2.85 per 1,000 a typical agent run costs a fraction of a cent, but an unbounded retry loop is still an unbounded retry loop.

Speed matters more in an agent loop

A multi-step agent may hit several challenges in one run, and each one sits on the critical path. Our median is 4.8 seconds on reCAPTCHA v2, measured on our own infrastructure — see the solver pages for per-type figures.

Ready to build?

Start solving for free and scale per solve as your traffic grows. No credit card needed.