[RECAP] Apollo Community Meetup - AI Agents Meet the Graph: Exploring the Model Context Protocol

Events and User Groups Community Meetups
, , ,
1

:apollo_logo_orange: :busts_in_silhouette: [RECAP] AI Agents Meet the Graph: Exploring the Model Context Protocol

:date: Meetup Date: April 30, 2025
:studio_microphone: Featured Speaker: @Dan_Boerner Apollo Graph Champion

:brain: MEETUP RECAP

This week, over 130 community members joined us to explore one of the hottest topics in APIs right now: AI agents as clients of our APIs :robot:. We focused on the Model Context Protocol (MCP), which is quickly gaining attention as a way to expose, connect, and integrate APIs as tools—alongside other tools and resources :hammer_and_wrench:.

I opened with some framing context, then @watson demoed the upcoming Apollo MCP Server, showing how Claude Desktop could answer questions about an internal graph. It got people thinking: what’s possible when agents meet the graph? :thinking:

Then the discussion took off. A few highlights:

  • Schema introspection vs. schema slices :fork_and_knife:: Should agents get dynamic access to the full schema or a curated subset? During prototyping, introspection can be useful—especially on smaller graphs—but latency and unpredictability grow with graph size. As Valerie England (Continued) noted, more tools ≠ better if you’re aiming for fast, consistent responses .
  • Real-world adoption stories :megaphone:: @Jonathan_Wondrusch (Block) and @scottsilvi (Team Duality) shared how they’re building with MCP. Jonathan offered a compelling case for dynamic introspection—even with its tradeoffs. Check the recording for his take :movie_camera:.
  • Flexibility vs. determinism :control_knobs:: MCP’s flexibility is powerful, but we talked about the value of more structured, predictable ways to expose schema functionality. A single approach may not fit every use case.
  • Internal use cases shine :sparkles:: Agentic AI as an internal design-time co-pilot stood out as a near-term win. Think schema reviews, query design helpers, and Slack-integrated API discovery—all driven by natural language and grounded in your actual graph :speech_balloon::bar_chart:.

There was so much more, but I’ll stop there. Catch the recording, share your thoughts below :down_arrow:, and get ready—this won’t be our last conversation about AI and MCP.

~ Dan

View Full Transcript

:pushpin: MEETING RESOURCES

:link: Here are some resources that were shared during the talk:

2 Likes