MCP over HTTP Proxy fuer Cursor-Tools - Integration mit Open WebUI und n8n
Configuration
View docs{
"mcpServers": {
"bbajor-mcpo": {
"url": "http://localhost:8000",
"headers": {
"MCPO_PORT": "8000",
"MCPO_API_KEY": "dein-geheimer-api-key-hier",
"WORKSPACE_PATH": "D:\\workspace\\pvs",
"MCP_SERVER_NAME": "cursor-mcp",
"MCP_SERVER_VERSION": "1.0.0"
}
}
}
}You run an MCP (Model Context Protocol) server that exposes Cursor tools over an HTTP API. This lets you query codebase content, read files, and perform workspace operations from automation tools like Open WebUI and n8n, while benefiting from a centralized, API-driven workflow for code analysis and manipulation.
Start by running the MCP server so you can access its HTTP API at the base URL. After the server is up, you can call endpoints to search the codebase, read files, or list directories. Open WebUI can connect to the server to configure and enable tools, and you can trigger the same capabilities from n8n using standard HTTP requests.
Key endpoints are exposed under the base URL and documented in the OpenAPI specification. Typical usage patterns include performing a semantic code search, fetching file contents from your workspace, or enumerating workspace directories to drive automated workflows. You can also verify the server is healthy and ready by checking the health endpoint.
Prerequisites you need installed before you begin:
- Node.js (for local builds and MCP server execution)
- Podman (or Docker) for containerized deployment
- PowerShell (on Windows) if you plan to use the Windows installation script
Follow these steps to install and start the MCP server using the manual flow (no Windows script):
npm install
cp .env.example .env
# Edit .env to set your API key and other config values
podman-compose up -d
podman-compose ps
# MCP server should be available at http://localhost:8000Environment variables you will typically configure in your .env file include port, API keys, and server identity. For example, you set the port the MCP over HTTP proxy listens on and the API key used to secure access. The workspace path is mounted as a volume so the MCP server can access the code and files you want to work with.
Open WebUI integration and the n8n workflow example assume the MCP server is reachable at http://localhost:8000 and that you provide the API key to authorize requests.
If the port is already in use, determine what process is bound to 8000 and either stop it or move the MCP server to another port in your .env file, then restart the containers.
If Podman commands are not available on your system, install Podman or switch to an alternative container runtime supported by your environment.
Check logs for the MCP server and the container to diagnose startup issues. You can access container logs with the standard Podman tooling and verify that the server process is running as expected.
Open WebUI: add an MCP Server connection via Admin → Settings → External Tools. Use the following values: Type: MCP (Streamable HTTP), URL: http://localhost:8000, API Key: your MCPO_API_KEY value. Enable the available tools and save the configuration to start querying the MCP server from the UI.
n8n: trigger actions by sending HTTP requests to the MCP server’s endpoints. For example, to perform a codebase search, send a POST to the codebase_search endpoint with the appropriate payload and your API key in the Authorization header.
The MCP server provides a set of tools that you can invoke over HTTP to perform actions such as searching the codebase, reading files, and listing directories within your workspace.
Semantically searches the codebase to locate relevant code fragments or concepts based on your query.
Reads a file from the configured workspace to retrieve its contents.
Lists the contents of a directory within the workspace, optionally in a recursive manner.