Hi Tip-Sheeters,

This week, I'm sharing details on an upcoming virtual event I'm speaking at, which is hosted by Mike Admundsen, author of multiple books on APIs. I also have a conversation with Mike later in the newsletter.

The Upcoming Event

I'd like to invite you to the API Superstream: AI and Agents on the O'Reilly Platform, July 17 from 10 am - 2 pm Central time.

Here are a few of the planned presentations:

• Building and Securing Remote MCP Servers Directly from Your API Gateway – Martyn Davies
• Empowering Your Agents: APIs as Tools with Azure API Management and MCP – Julia Kasper
• Bridging APIs and AI: LLMs, GenAI, and the Future of XAI Systems – Manai Mohamed Mortadha
• Using APIs with AI – Ryan Day

How to Attend

The workshop is for members of the O'Reilly Platform (which has a monthly fee). Here's how you can attend:

Log in and sign up for the superstream: https://learning.oreilly.com/live-events/api-superstream-apis-and-agents/0642572173432/0642572173425/​

A Conversation With Mike Admundsen

Since Mike's organizing the Superstream, I took time to catch up with his thoughts on the current goings-on with APIs. Mike A is a well-known API expert and the author of multiple books with O'Reilly and Pragmatic Programmers. Mike shared some of his thoughts about the intersection of AI and APIs.

Ryan: AI applications and Agents are beginning to consume APIs. How do you see this impacting the way that API producers design and manage their APIs?

Mike: For years, API design advice has been telling people to build APIs that are well-documented, provide clear workflows, and contain lots of rich metadata to help API consumers better understand how the API works and how they can take advantage of that in creating powerful API client applications.

Now, with LLM-driven API consumers, all that advice comes up again. Not only do AI-powered API clients need rich metadata, clear flows, and helpful documentation, these machine clients need help making good choices on which API path to take and what inputs to supply in each interaction. I think the rise of AI-driven APIs gives us all another chance to lean in on creating well-designed, well-documented (both at design time and runtime) services that are resilient, reliable, and scalable.

Ryan: What are some critical skills that individuals in the API development and management space can benefit from right now?

Mike: I tell my customers to focus on making their API interactions observable and measurable. Now that we have a new type of API consumer out there, we need to pay close attention to how well these clients are doing when it comes to discovering, navigating, and interacting with our APIs. Do we see lots of failed attempts? That’s likely telling us something about how to better design the API. Do we see some API workflows consistently returning positive traffic and others ending up creating more support tickets and development time? Then maybe we should focus on strengthening the positive flows and reducing the ones that don’t pay off for customers.

This also leads to increased needs for building machine-level API discovery tools (not just catalogs) and more consistent descriptions of input arguments and return payloads. AI clients are great at noticing patterns and repeating them. They can easily find loose associations between variable names (first_name, givenName, fname, etc.) and are more than happy to make assumptions about how to supply required inputs based on these associations. We should take advantage of that skill by supplying API bots with extensive documentation, clear flows, and consistent naming. This is another level of API design management, a skill that I think we’ll need to improve over the near term.

Ryan: You are facilitating the Superstream on July 17. What do you hope that the audience will get from attending this event?

Mike: This is a chance for attendees to hear from experienced professionals working at the conjunction of APIs and AI and I hope everyone goes away with a bit more understanding of how the current crop of generative AI tools operate and how we can leverage that in the designing and implementing of APIs. There is so much to cover, I suspect most people will see this as a set of pointers from great minds leading to further study by all parties.

Keep Coding,

Ryan Day

👉 https://tips.handsonapibook.com/ -- no spam, just a short email every week.