Follow-along: Generate a graph (⏳ 25 min)

Goal: Use an agent and mcp tools to orchestrate two REST APIs into one graphql schema.

Create the schema for Space Devs

We've provided a JSON file with the Space Devs Launch Library API endpoint, located in the LaunchLibraryAPI.json file in the local project.

1. Open up goose and enter the following prompt to create the schema for Space Devs using the LaunchLibraryAPI.json file.

   I just started a new graph with Apollo, I'm providing the endpoint for the Space Devs Launch Library at LaunchLibraryAPI.json in the local folder. I want to get the only the important details with fields for launches. Pad information must be included with lat/long values.

   Copy

2. The agent will prompt you to confirm the plan. It will commonly ask if you want the use the spacedevs development endpoint and if you want your schema in one file. Answer yes to both.

   Troubleshooting: Running rover dev

   If the agent does not automatically run the project, open a new terminal window and run the following command.

   rover dev --supergraph-config supergraph.yaml

   Copy

   Alternatively, you may choose to prompt goose to run the dev server for you.

   Run the project locally.

   Copy

3. Navigate to http://localhost:4000/ where the Explorer is running.

   Troubleshooting: Schema errors after goose is finished

   If you notice an error that goose did not address, prompt goose to continue. For example:

   There are schema composition errors. Please address these and test again.

   Copy

4. Create a query using the + buttons in the Documentation panel.

   Below is an example, the agent may have structured your schema slightly differently.

   

   Example query

   query UpcomingLaunches($limit: Int) {
     upcomingLaunches(limit: $limit) {
       id
       name
       pad {
         latitude
         longitude
       }
     }
   }

   Copy

Integrate Open Weather API

1. In goose, enter the following prompt to integrate the Open Weather API into the schema.

   Add a field-level connector on Pad.weather. Use the open weather endpoint and the OPEN_WEATHER_KEY in the env file to research and understand this API https://api.openweathermap.org/data/2.5/weather?lat={lat}&lon={lon}&appid={OPEN_WEATHER_KEY}

   Copy

   ⚠️ Warning: Using .env files when working with agents can expose your keys. This flow is not to be used for sensitive data.

2. Navigate back to http://localhost:4000/ and your new API fields will have hot-reloaded.

3. Add weather to the query and click Run again to test.

   Example query

   query UpcomingLaunches($limit: Int) {
     upcomingLaunches(limit: $limit) {
       id
       name
       pad {
         weather {
           visibility
         }
       }
     }
   }

   Copy

Explore the graph

1. In the http://localhost:4000/ Sandbox, click on the Schema icon in the left hand navigation.

2. Click on the Visualization tab. This view allows you to see the graph view of your schema structure including how everything is connected. This is another tool that can help you validate your agentic output.

   

💭 Tips

This project has an Agents.md and a .goosehints file that help us give our LLM a "persona" on building a graph using Apollo Connectors. Prompting with words like "Generate me a connector for..." help increase the vector search hits with our GraphOS MCP tools.

Here are some tips we used in creating these prompts: