Building GraphRAG Python MCP tools

[adam-cowley]

Caption: Build your own GraphRAG MCP server with graph-backed tools and resources.

Pre-Submission Checklist

Content Creation

• Write course content (modules, lessons, questions, challenges)
• Add any required images and includes to lesson folders
• Set course status to :status: draft in course.adoc
• Follow existing course structure (used recent course as template)
• Include proper course metadata (categories, duration, etc.)
• Create course overview in course.adoc
• Create summary in summary.adoc

Assets & Media

• Create course illustration/badge (discuss with Adam if needed)
• Create banner image
• Ensure all media assets are optimized

Testing & Quality Assurance

• Run unit tests (npm run test)
• Test course locally in development environment
• Verify all links work
• Check Cypher queries and code samples are valid (if applicable)
• Ensure the course can be deployed without errors
• Check display includes correctly in the development environment
• Ensured course follows GraphAcademy content standards (Cursor rules)

[martinohanlon]

I think the use of MCP Inspector is a challenge. I really struggled to get it working in a container. My view would be to either simplify and accept that the learner has to create their own development environment or look for / build a test bed so learners can "have a go".

I agree that using trying to test / develop the tools through an agent interface would be really difficult. I think howegver there should be a lesson towards the end where the learner uses the tools they have built as part of the [VS Code] agent.

[martinohanlon]

Is there a practical / challenge for prompts?

[adam-cowley]

I'm struggling with how to test this in the client, will have a think...