Berghain Events MCP Server

Berghain Events API & MCP Implementation

Project structure

Requirements

• Python 3.10+
• uv (for dependency management and virtual environment)
• AWS Account (for DynamoDB and App Runner)
• Firecrawl API Key

Facility

1. Clone this repository:
   git clone <repository-url> cd berghain-api
2. Create and activate a virtual environment with uv :
   uv venv source .venv/bin/activate # En Linux/macOS # .venv\\Scripts\\activate # En Windows
3. Install dependencies with uv :
   uv pip install -r requirements.txt

Detailed Process

1. Data Extraction with Firecrawl

• Configure the Firecrawl MCP in Cursor: Make sure you have your Firecrawl API Key. In the Cursor MCP settings, add:
  "firecrawl-mcp": { "command": "npx", "args": [ "-y", "firecrawl-mcp" ], "env": { "FIRECRAWL_API_KEY": "fc-YOUR_FIRECRAWL_API_KEY" } }
• Run the extraction: Use an agent in Cursor (or a similar tool) to call the Firecrawl MCP and ask it to extract events from https://www.berghain.berlin/en/program/ .
• Save the data: The extracted data should be saved as JSON files in the events/ directory. For example, events/berghain_events_YYYY-MM-DD.json .

2. Loading Data into DynamoDB

• Create the table in DynamoDB: The scripts/create_table.py script takes care of this. Run it (adjust the parameters if necessary):
  uv run python scripts/create_table.py --table berghain_events --region tu-region-aws # Para desarrollo local con DynamoDB Local (ej. docker run -p 8000:8000 amazon/dynamodb-local): # uv run python scripts/create_table.py --table berghain_events --region localhost --endpoint-url http://localhost:8000
• Load data into the table: The scripts/load_data.py script loads events from JSON files.
  uv run python scripts/load_data.py --table berghain_events --region tu-region-aws --path events # Para desarrollo local: # uv run python scripts/load_data.py --table berghain_events --region localhost --endpoint-url http://localhost:8000 --path events
  Make sure that app/core/config.py (imported by load_data.py ) has the necessary configurations if you don't pass them as arguments.

3. API with FastAPI and MCP Server

• The API logic resides in the app/ directory, with the endpoints defined (e.g., in app/api/endpoints/events.py ).
• The app/main.py file is configured to launch the FastAPI application and the FastMCP server, exposing the API endpoints as tools for AI models. Review the custom_maps in app/main.py to see how GET routes are mapped to RouteType.TOOL .

Deployment on AWS

a. Dockerfile

Make sure your Dockerfile is configured correctly to use uv and run app/main.py :

b. Build and Upload Image to Amazon ECR (Elastic Container Registry)

1. Authenticate Docker with ECR:
   aws ecr get-login-password --region tu-region-aws | docker login --username AWS --password-stdin tu-aws-account-id.dkr.ecr.tu-region-aws.amazonaws.com
2. Create a repository in ECR (if it doesn't exist):
   aws ecr create-repository --repository-name berghain-mcp-api --region tu-region-aws
3. Build your Docker image:
   docker build -t berghain-mcp-api .
4. Tag your image:
   docker tag berghain-mcp-api:latest tu-aws-account-id.dkr.ecr.tu-region-aws.amazonaws.com/berghain-mcp-api:latest
5. Upload the image to ECR:
   docker push tu-aws-account-id.dkr.ecr.tu-region-aws.amazonaws.com/berghain-mcp-api:latest
   Replace tu-region-aws and tu-aws-account-id with your values.

c. Deploy Infrastructure with Terraform

1. Prepare your Terraform files: Make sure you have your Terraform configuration files (e.g., main.tf , variables.tf , outputs.tf ) in a directory (e.g., terraform/ ). These files should define the necessary AWS resources, such as the AWS App Runner service that will use the ECR image, and the DynamoDB table (if it is also managed by Terraform). Your App Runner configuration in Terraform should reference the uploaded ECR image.
2. Navigate to the Terraform directory:
   cd terraform
3. Initialize Terraform:
   terraform init
4. Apply the Terraform configuration:
   terraform apply
   Review the plan and confirm the application. Terraform will provision the resources.
5. Get the service URL: Once applied, Terraform should display the defined outputs, including the App Runner service URL. Make a note of this URL (e.g., https://<id-servicio>.<region>.awsapprunner.com ).

Test the Deployed Solution

a. Adjust Local Test Script ( mcp_local.py )

Set the mcp_local.py file to the root of your project

Important: Update the mcp_server_url variable in mcp_local.py with the URL you got from the Terraform output.

b. Run Test

From the root of your project (or wherever you saved mcp_local.py ):

This will run the PydanticAI agent, which will attempt to connect to your deployed MCP and perform the query.

License

MIT