How to Install CapSolver MCP from the Official MCP Registry

TL;DR

• CapSolver MCP is listed in the Official MCP Registry as io.github.capsolver-ai/capsolver-mcp. The active latest release is 0.1.3, distributed through PyPI and launched locally over stdio.
• The shortest installation path is uvx capsolver-mcp. uvx creates an isolated environment and runs the published command without requiring a permanent global package installation.
• Every client configuration needs CAPSOLVER_API_KEY. Keep the real key in the client environment or an approved secret store, never in a public repository.
• A working connection should expose five tools. They cover token-based solving, page detection, page-level solving, balance checks, and supported challenge types.
• Browser-based tools require the optional browser extra and Playwright Chromium. Basic token-based tools do not need that additional browser installation.
• Use CapSolver only in workflows you own or are authorized to test or automate. Registry availability does not grant permission to access private or restricted systems.

Install CapSolver MCP from the Official MCP Registry

The Official MCP Registry gives MCP clients and developers a canonical record for discovering CapSolver MCP, its package, transport, required environment variable, repository, and current version. The registry does not host the Python package; it points clients to the verified capsolver-mcp release on PyPI.

This tutorial walks through discovery, installation, configuration, and the first connection check. It uses CapSolver only for lawful, authorized AI-agent, QA, RPA, and browser-automation workflows.

What the Official Registry Entry Provides

The CapSolver MCP entry publishes the exact metadata a compatible client needs to construct a local stdio configuration.

You can inspect the current entry in the Official MCP Registry search interface. The registry is an upstream catalog: it standardizes discovery metadata while the actual package remains on PyPI. The MCP Registry launch documentation describes this catalog-and-distribution model.

Prerequisites

CapSolver MCP needs Python 3.10 or newer, a compatible MCP client, and a CapSolver API key.

Prepare the following before editing a client configuration:

1. Install Python 3.10 or newer.
2. Install uv if you want the recommended uvx workflow.
3. Create a CapSolver account and obtain an API key from the dashboard.
4. Confirm that your MCP client supports local stdio servers.
5. Decide whether you need only token-based tools or the optional browser tools.

The CapSolver MCP service documentation explains the supported integration path, while the CapSolver AI agent overview helps you choose between MCP, agent tools, and the Core SDK.

Step 1: Find the Correct Registry Record

Search for capsolver in the Official MCP Registry and select this exact name:

io.github.capsolver-ai/capsolver-mcp

Confirm that the displayed version is 0.1.3 or newer and that the status is active. Avoid similarly named community packages unless you have separately reviewed their publisher, repository, permissions, and source code.

For an API-level check, request the exact server version:

curl "https://registry.modelcontextprotocol.io/v0.1/servers/io.github.capsolver-ai%2Fcapsolver-mcp/versions/0.1.3"

The slash in the registry name is encoded as %2F. The response should identify capsolver-mcp as a PyPI package, uvx as the runtime hint, and stdio as the transport.

Step 2: Install and Verify the CLI

The recommended path runs CapSolver MCP in an isolated environment:

uvx capsolver-mcp --help

If you prefer a managed virtual environment, install the package explicitly:

python -m pip install "capsolver-mcp==0.1.3"
capsolver-mcp --help

The current CLI supports stdio, sse, and streamable-http runtime modes. The Official MCP Registry entry uses stdio because clients launch the local package as a subprocess. The package can also start an HTTP-capable process on your own infrastructure, but CapSolver does not currently publish a hosted remote MCP endpoint in the registry.

The uvx approach follows the isolated tool execution model documented in the official uv tool guide.

Redeem Your CapSolver Bonus Code

Boost your automation budget instantly!
Use bonus code CAP26 when topping up your CapSolver account to get an extra 5% bonus on every recharge — with no limits.
Redeem it now in your CapSolver Dashboard

Step 3: Configure a Local MCP Client

A local client must launch uvx capsolver-mcp and pass CAPSOLVER_API_KEY to the subprocess.

Use this configuration shape in clients that accept an mcpServers object, including Claude Desktop and clients with compatible local-server configuration:

{
  "mcpServers": {
    "capsolver": {
      "command": "uvx",
      "args": ["capsolver-mcp"],
      "env": {
        "CAPSOLVER_API_KEY": "YOUR_CAPSOLVER_API_KEY"
      }
    }
  }
}

Replace the placeholder locally. Do not commit a real API key to Git, paste it into issue reports, or include it in screenshots.

If the desktop application cannot find uvx, use the absolute path returned by the following command:

which uvx

On Windows PowerShell, use:

Get-Command uvx

After saving the configuration, fully restart the MCP client. A window close may leave the application running in the background, so quit the process if the server does not appear after editing the configuration.

Step 4: Verify the MCP Handshake and Tools

A successful connection should expose five CapSolver MCP tools to the client.

Start with a non-browser check such as:

Use the CapSolver MCP server to list the supported CAPTCHA types.

Then verify account authentication with:

Use the CapSolver MCP server to check my account balance.

These checks confirm that the client can launch the stdio process, discover its tools, and pass the environment variable. A real solving request additionally needs authorized target parameters and sufficient account balance.

Step 5: Enable Browser-Based Tools When Needed

Browser-based detection and page interaction require the optional package extra plus Playwright Chromium.

For a dedicated environment, run:

python -m pip install "capsolver-mcp[browser]==0.1.3"
playwright install chromium

If the MCP client launches uvx, include the browser extra in the package argument:

{
  "mcpServers": {
    "capsolver-browser": {
      "command": "uvx",
      "args": ["--from", "capsolver-mcp[browser]==0.1.3", "capsolver-mcp"],
      "env": {
        "CAPSOLVER_API_KEY": "YOUR_CAPSOLVER_API_KEY"
      }
    }
  }
}

Playwright still needs a compatible Chromium binary on the same machine. For production workflows, pin versions, test upgrades in staging, and keep browser installation separate from application secrets.

Common Setup Problems

Most CapSolver MCP connection failures come from the runtime path, Python version, missing credentials, or optional browser dependencies.

The client reports spawn uvx ENOENT

The desktop application cannot see uvx on its process PATH. Use the absolute uvx path in the command field, then restart the application completely.

The server starts but tools fail authentication

The CAPSOLVER_API_KEY value is missing, malformed, or not reaching the child process. Confirm the variable name exactly, replace the placeholder, and avoid adding extra quotation marks inside JSON values.

detect_captchas or solve_on_page is unavailable

The base package does not install Playwright. Install capsolver-mcp[browser], install Chromium, and restart the client so it launches the updated environment.

The Registry shows more than one CapSolver result

Registry search APIs can return multiple published versions of the same server. Select the record where isLatest is true, or query the exact version endpoint rather than treating the result count as a uniqueness check.

The client connects but a solve request fails

Separate connection problems from task problems. First verify tool discovery and get_balance; then check the task type, website URL, site key, optional fields, and account balance against the CapSolver documentation.

Operational and Responsible-Use Limits

CapSolver MCP is a local tool interface, not permission to automate a target. Use it only on websites and systems you own or are authorized to test or automate. Follow applicable terms, privacy requirements, rate limits, and data-retention rules.

The Official Registry confirms publisher-controlled metadata and package location; it does not replace your own security review. Inspect the CapSolver MCP source repository, pin versions where your change process requires it, protect API keys, and test updates before broad deployment.

For architecture decisions beyond a single client, see how enterprise AI-agent teams structure authorized challenge handling and review the definition of an API endpoint when designing service boundaries.

Conclusion

The Official MCP Registry now provides a canonical, searchable record for CapSolver MCP, while PyPI supplies the executable Python package. A reliable setup follows a short sequence: confirm the registry identity, run the CLI, add CAPSOLVER_API_KEY to the client environment, restart the client, and verify tool discovery before attempting an authorized task.

Use CapSolver to add documented CAPTCHA-solving tools to supported AI-agent and browser-automation workflows.

Build Reliable Automation Workflows with CapSolver

Create an account, store your API key securely, and start with a bounded test on a website or system you are authorized to automate. The CapSolver FAQ covers common product and account questions before you move the integration into production.

FAQ

Q: Is CapSolver MCP available in the Official MCP Registry?

Yes. The registry name is io.github.capsolver-ai/capsolver-mcp, and version 0.1.3 is currently active and marked as the latest release.

Q: Does installing from the Registry create a hosted remote connection?

No. The current registry entry describes a PyPI package launched locally with uvx over stdio; it does not advertise a CapSolver-hosted remote MCP endpoint.

Q: Do I need to install capsolver-mcp globally?

No. uvx capsolver-mcp runs the tool in an isolated environment, while an explicit pip install remains available for teams that manage their own virtual environments.

Q: Why does my client connect but not show the browser tools?

The browser tools require the capsolver-mcp[browser] extra and a Playwright Chromium installation. Restart the client after installing both components.

Q: Which environment variable does CapSolver MCP require?

CapSolver MCP requires CAPSOLVER_API_KEY. Pass it through the client environment and keep the real value out of source control and support screenshots.

Q: Can I use CapSolver MCP on any website?

No. Use CapSolver MCP only on websites and systems you own or are authorized to test or automate, and follow the applicable terms, privacy rules, and legal requirements.