Notes from the runtime.
How we build, govern and price agents that run in production, plus what we learn shipping them with customers.
Every article, filed by topic.
- GovernanceRead article
Policies that travel with the action
A rule written into a prompt can be talked around. A rule attached to the action itself cannot. How Ordin's policy engine decides every call.
- ProductRead article
What a finished run really costs
Per-token pricing hides the cost of retries, dead ends and abandoned jobs. We switched to pricing per finished run, and here is the maths behind it.
- EngineeringRead article
Shared memory without the shared mess
Six specialists, one job, one memory. How Ordin keeps agents in sync without letting one agent's scratch notes leak into another's work.
- GovernanceRead article
A human in the loop that actually scales
Sign-off queues usually turn into a bottleneck within a week. Three design choices that keep people in control without making them the slowest step.
- Field notesRead article
Field notes from our first hundred agents
What we learned shipping a hundred agents with customers: the boring problems, the surprising ones, and the checklist we now start every project with.
Hand it the job. Keep the rules.
Free for the first 10,000 runs. Bring your own model, keep your own data, ship your first agent this afternoon.