Run with Docker

Clone the Repository

git clone [email protected]:lensesio/lenses-mcp.git

Build and Tag Image

Build and tag a Docker image using the Dockerfile included at the root of the repository.

docker build -t lenses-mcp:local

Run using Docker

Run using Docker by passing in configuration as environment variables for the default setup that uses OAuth for authentication and authorization. This example runs the MCP server using HTTP transport, available on port 8000 on the local machine. This is set in the MCP_ADVERTISED_URL variable.

docker run --rm -it --name lenses-mcp \
-p 8000:8000 \
-e TRANSPORT=http \
-e HOST=0.0.0.0 \
-e PORT=8000 \
-e OAUTH_ENABLED=True \
-e LENSES_URL=https://your-lenses.io \
-e LENSES_ADVERTISED_URL=https://your-publicly-available-lenses.io \
-e MCP_ADVERTISED_URL=http://localhost:8000 \
lenses-mcp:local

Configure in your AI tool

To run in Cursor, Claude, Codex, etc. use the following mcp.json configuration to configure the client to connect using HTTP transport.

{
  "mcpServers": {
    "Lenses": {
      "type": "http",
      "url": "http://localhost:8000/mcp"
    }
}

Authenticate with OAuth

Navigate to your AI tool's MCP server settings and connect to authenticate with the Lenses OAuth authorization server (visible to your AI tool's client and configured with LENSES_ADVERTISED_URL). Here is an example using Cursor.

This opens a web browser at LENSES_ADVERTISED_URL/oauth2/authorize and passes OAuth parameters. Log in with your username/password or SSO (example shown is the Community Edition screen).

You will be presented with an OAuth authorization consent screen where you can set read, write and delete scopes.

Completing this flow should take you back to your AI tool, authenticated and ready to use Lenses MCP Server with the granted authorization scopes.