How Orrery handles it
Budget per client. Proof per task.
Orrery already proposes every change as a diff you approve, runs your tests, and writes a receipt for each run. The agency layer sorts that by client.
Available now
Per-client cost caps
Set a hard spend limit for each client. The agent stops at the cap instead of running over, so a fixed bid stays fixed.
Available now
Client receipts
Export a receipt for any task: what changed, which tests passed, preview screenshots, the model used and what it cost. Attach it to an invoice or a pull request.
Available now
Monthly AI-use log
One log per client per month: tasks, models, and where each one ran. A starting point for the AI-disclosure section of a client report.
Available now
Your keys or local models
Use your own provider keys, or run models on your own hardware (Ollama, LM Studio, vLLM), so client code can stay on your machines when the contract asks for it.
Available now
Hosted models included
Plans include hosted-model credits, so you can start without setting up a provider account. Credits stop hard at the limit and never overage.
Available now
Diffs you approve
The agent never writes straight to your repo. Every change is a reviewable diff, with rewind if a step goes wrong, so you can stand behind what you ship.
All of this is in Orrery today, from version 4.0.0-rc.51. Set a client up in Settings → Workspaces → Client.