Arazzo vs Traditional Chatbots: What Actually Works?

What happens when you give an AI agent hundreds of API endpoints and hope it figures out the right workflow? Spoiler: it nearly gets it right... but never reliably. In this talk, Frank Kilcommins (Head of Enterprise Architecture at Jentik and co-author of the Arazzo Specification) breaks down why API documentation quality is the core knowledge problem holding agentic systems back (and how Arazzo solves it).

Arazzo (Italian for "tapestry") is an open standard under the OpenAPI Initiative that lets developers describe deterministic, use-case-oriented API workflows in JSON or YAML. Instead of handing an LLM a sprawling OpenAPI doc with thousands of endpoints, Arazzo gives agents proven recipes for success. Step by step, with success criteria, retry logic, and branching built in.

In this session you'll learn:

What separates agent experience from developer experience (and why both matter)
Why poor API documentation leads to wrong tool calls, bad arguments, and silent failures
How Arazzo workflows complement MCP with just-in-time tool delivery
Why deterministic execution beats LLM reasoning for known, critical workflows
How sandboxed environments can help you safely discover and codify new agent workflows

Whether you're building agentic products, designing APIs for the AI era, or just trying to get your agents out of hello-world territory and into production -- this talk is essential viewing.

GitKraken Desktop:
http://tr.ee/GKDYT

GitKraken CLI:
http://tr.ee/CLIYT

GitLens for VS Code:
http://tr.ee/GLYT

Git Integration for Jira:
http://tr.ee/GijYT

Git Blog:
http://gitkraken.com/blog