Put Governance Where the Agent Can Reach It

I made the case last week that you should govern in the IDE, where the work happens — get the ruleset in front of the engineer while they’re still typing, instead of waiting for a red CI build to tell them what they already forgot. That argument has a hole in it now, and the hole is the part of the work that isn’t typed by an engineer at all. An agent drafting an OpenAPI document in a chat, wiring up an Arazzo workflow on request, scaffolding an MCP server for you — none of that touches an IDE, and none of it touches CI until much later, if ever.