investigate MCP support

[dietervu]

As someone suggested to @Querela during the annual conference 2025: check if we can somehow provide an endoint for MCP that exposes the results the aggregator has collected from the endpoints.

Related:

• MCP for the VLO: investigate MCP support VLO#405
• MCP for the CLARIN website: Look into options to expose drupal content proactively to AI bots CLARIN-website-tracker#10

[Querela]

Resources:

• https://product.hubspot.com/blog/unlocking-deep-research-crm-connector-for-chatgpt (!)
• https://medium.com/@cstroliadavis/building-mcp-servers-536969d27809

[Querela]

Possible MCP Resources:

• List of resources (in the FCS) containing resource ID, resource title and description, resource institution, other metadata like language and query languages
• Single resource details?
• Scan statistics (e.g., error reports), as well as Search statistics (errors mostly)

Tools?

• TODO: Search with results, streaming (SSE?) → support blocking search method or SSE streaming; also add better rate-limiting