Deep

Blog

Short posts that explain the plane to people who already run agents. This page is the whole publication — not a staging area for somewhere else. If you find this post through a search engine, a link from a friend, or a stray tab, you're already looking at the real thing.

Posts

• Why your agents need a control plane — the moment one script, one .env, and one deploy stops being enough.
• Two protocols, one plane — why Agent Protocol clients and Runner Protocol workers never talk to each other directly, and how that split is what lets a LangGraph agent, a CrewAI crew, and a hand-rolled Python script share one control plane without any of them knowing the others exist.
• BYO runner, governed secrets — your agent code and your compute stay exactly where they already run; what changes is how credentials reach them, and for how long they're allowed to work.
• Spend without becoming a billco — why "FinOps" here means budgets, holds, and an honest cost estimate, not a second Stripe integration bolted onto your agent platform.

There's no publishing calendar and no editorial backlog beyond what's linked above — each post gets written when there's something specific and true to say, not on a schedule. Related reading if you want more depth than a blog post gives: the chapters walk through the architecture more methodically, and the decision write-ups explain the specific engineering trade-offs behind choices these posts only gesture at. If you'd rather just see the plane running than read about it, the five-minute try is faster than either.