The project README describes idp-core as "instantly accessible to AI Agents via MCP" — but I couldn't find any MCP server implementation in the codebase. The src/ tree under com/decathlon/idp_core/ has no mcp package.
Why this matters: The Model Context Protocol (MCP) lets AI coding assistants (Claude Code, Copilot, Cursor, etc.) query and interact with tools directly. For an IDP backend, this means an AI assistant could:
- Query the software catalog for service ownership and dependencies
- Look up scorecards to check engineering excellence standards
- Trigger self-service actions (e.g., provision a new service) from a chat prompt
- Search entities with the existing filter/search DSL
What already exists that an MCP module could expose:
| Feature |
Existing code |
MCP tool idea |
| Entity catalog |
EntityService, EntityController |
catalog_search, catalog_get_entity |
| Scorecards |
docs/src/features/scorecards.md |
scorecard_get, scorecard_list |
| Self-service actions |
docs/src/features/self-service-actions.md |
actions_trigger, actions_list |
| Entity templates |
EntityTemplateService, EntityTemplateController |
template_list, template_get |
Suggested approach:
- Create a new
infrastructure/adapters/mcp/ package (following the existing hexagonal architecture pattern in infrastructure/adapters/api/)
- Implement a Spring Boot MCP server using the
spring-ai-mcp-server SDK or the official MCP Java SDK
- Expose read-only tools first (catalog search, scorecard lookup) — safe to deploy alongside existing REST API
- Add integration tests following the pattern in
src/test/java/com/decathlon/idp_core/AbstractIntegrationTest.java
Evidence: The hexagonal architecture is already in place — REST controllers live under infrastructure/adapters/api/controller/, so an MCP adapter under infrastructure/adapters/mcp/ follows the same pattern exactly.
The project README describes idp-core as "instantly accessible to AI Agents via MCP" — but I couldn't find any MCP server implementation in the codebase. The
src/tree undercom/decathlon/idp_core/has nomcppackage.Why this matters: The Model Context Protocol (MCP) lets AI coding assistants (Claude Code, Copilot, Cursor, etc.) query and interact with tools directly. For an IDP backend, this means an AI assistant could:
What already exists that an MCP module could expose:
EntityService,EntityControllercatalog_search,catalog_get_entitydocs/src/features/scorecards.mdscorecard_get,scorecard_listdocs/src/features/self-service-actions.mdactions_trigger,actions_listEntityTemplateService,EntityTemplateControllertemplate_list,template_getSuggested approach:
infrastructure/adapters/mcp/package (following the existing hexagonal architecture pattern ininfrastructure/adapters/api/)spring-ai-mcp-serverSDK or the official MCP Java SDKsrc/test/java/com/decathlon/idp_core/AbstractIntegrationTest.javaEvidence: The hexagonal architecture is already in place — REST controllers live under
infrastructure/adapters/api/controller/, so an MCP adapter underinfrastructure/adapters/mcp/follows the same pattern exactly.