Octoparse MCP Server

English | 日本語

Octoparse MCP Server connects MCP-compatible AI clients to Octoparse so users can discover scraping templates, create cloud tasks from Octoparse preset templates, manage existing tasks, and export structured data.

Hosted Server

Use the hosted Streamable HTTP endpoint:

https://mcp.octoparse.com

Configure this as an MCP server endpoint, not as a REST API tool.

Authentication

The hosted server officially supports both OAuth and API key authentication:

• OAuth: the default option when the MCP client can open a browser authorization flow.
• API key: an optional authentication method for clients that can send custom HTTP headers.

For API key mode, send the key in the x-api-key header. Do not place real keys in source control, issue reports, screenshots, or shared logs.

Quick Start

Generic HTTP MCP configuration:

{
  "mcpServers": {
    "octoparse": {
      "type": "http",
      "url": "https://mcp.octoparse.com"
    }
  }
}

Claude OAuth / default authentication:

claude mcp add --transport http octoparse https://mcp.octoparse.com

Claude with API key:

claude mcp add --transport http octoparse https://mcp.octoparse.com --header "x-api-key: YOUR_API_KEY"

Codex CLI:

codex mcp add octoparse --url https://mcp.octoparse.com

Gemini CLI example configuration:

{
  "mcpServers": {
    "octoparse": {
      "httpUrl": "https://mcp.octoparse.com",
      "oauth": {
        "clientId": "Octoparse",
        "enabled": true
      }
    }
  }
}

After adding the server, complete authorization in the browser if your client prompts for OAuth.

More client examples are in Configuration.

Tools

See Tools for inputs, examples, and common failure cases.

Typical Workflow

search_templates -> execute_task -> export_data

Example user prompt:

Find a cloud-capable Octoparse template for Amazon product listings, run it for wireless earbuds, and export the result as CSV.

For an existing task, use:

search_tasks -> export_data

Scheduling, notifications, recurring jobs, and external automations are handled by the MCP client or another automation system. The MCP server exposes task and export tools; it does not provide a standalone scheduler.

Scope

Octoparse MCP Server can:

• Search the Octoparse template library.
• Create and run cloud-capable tasks from Octoparse preset templates.
• Validate template parameters before execution.
• Export collected data when a task has data available.
• Search, start, stop, monitor, and export existing cloud-supported tasks, including existing cloud-supported custom tasks.

It does not:

• Run local-only desktop tasks through the hosted server.
• Create fully custom scraping tasks.
• Create or edit custom task configuration details such as custom fields, pagination rules, loop logic, browser workflow, or extraction steps.
• Upload private templates.
• Provide billing or account administration tools.
• Schedule recurring runs by itself.

Common Limits

• Some templates are local-only and must be run in the Octoparse desktop app.
• Cloud execution depends on account permissions, available credits or quota, target-site availability, and template support.
• Long-running tasks may need follow-up calls. If data is still collecting or exporting, wait and call export_data again.
• Tool availability can be narrowed with includeTools or excludeTools URL parameters. See Overview.

Security

• Keep API keys and OAuth tokens out of Git, logs, screenshots, and public issues.
• Use OAuth when possible for interactive clients.
• Use x-api-key only with clients that store headers securely.
• Report vulnerabilities privately. See Security.

Documentation

• Overview
• Configuration
• Tools
• Troubleshooting
• Octoparse API Docs
• Help Center
• Available on Smithery

Development

npm install
npm run build
npm test

For local environment variables, copy .env.example to .env and fill only values needed for your environment. Never commit .env.

Contributing

Contributions are welcome. Please read CONTRIBUTING.md before opening issues or pull requests.

License

MIT. See LICENSE.md.