installation needs to have enabled [Service to Service](https://backstage.io/docs/auth/service-to-service-auth/) communication. Users need to have access of the token used for such communication. The `yaml` snippet below shows where and how it's setup. ```yaml app: # ... backend: # ... auth: # ... externalAccess: - type: static options: token: subject: curl-requests ``` ## Anatomy ### Dependencies The project is using: ```xml io.quarkiverse.mcp quarkus-mcp-server-stdio 1.0.0.Alpha5 ``` for implementing an mcp server that reads from stdin and writes to stdout. It also uses: ```xml io.quarkiverse.backstage quarkus-backstage 0.4.1 ``` For talking to the [Backstage](https://backstage.io) API. ### The implementation The implementation is pretty straight forward, as there is a tiny file needed: [[src/main/java/org/acme/Backstage.java]] ## Setting up goose [Goose](https://github.com/block/goose) is an local AI agent that runs as an interactive shell and is supports plugins (including mcp servers). To setup goose so that it uses this mcp server add the followng extension to your config.yaml: ```yaml quarkus-backstage-mcp: args: - --quiet - /home/iocanel/demo/backstage-mcp/target/quarkus-app/quarkus-run.jar cmd: jbang enabled: true envs: {} name: quarkus-backstage-mcp type: stdio ``` ### Sample prompts #### Listing the templates ```bash list all the available backstage templates ``` ### Instantiating a template To instantiate a template one needs values.yaml file containing the template parameters to use. The default values can be extracted from the template using the backstage CLI: ```sh quarkus backstage template info --show-default-values

#backstage

Backstage MCP

Bee Agent & Model Context Protocol with Backstage