Your agents do the work.
You keep control.
Orchestrator Zero (oz0) helps platform and IT teams run AI work on their own infrastructure. Give an agent a task, control its actions and spending, then inspect what happened.
Illustrative workflow using your configured tools, not a captured run.
Your request
“Why is checkout slow?”
- 01
Receive the request
- 02
Read the logs
- 03
Check the findings
- 04
Return a report
Result ready
You get findings to review, with the steps the agent took to reach them.
Where does oz0 fit?
Your app or team provides the task. oz0’s default runtime manages the agent’s model and tool loop, with the machines and limits you provide. External harnesses are optional.
- Your app or team Task ↓
- Your oz0 server Job ↓
- Agents and tools on your machines
One server binary runs management, the edge and embedded Temporal together by default. You can separate these roles or use external Temporal. Nodes use separate software; server and node can share a machine.
You operate the server and worker machines, including updates and backups. Choose compatible hosted or self-hosted models; serving them is separate from oz0. Providers and connected tools may receive the data their calls need.
| The agent | Chooses its next step using the tools and information available. |
|---|---|
| oz0 | Runs the default agent loop, applies relevant permissions, approvals and budgets, and records execution. |
| Your team | Chooses plugins and providers, grants access, sets limits and operates the deployment. |
Make one small task useful first.
1. Pick a real problem
Start by investigating one failed connection. Decide what evidence makes the answer useful.
2. Set its boundaries
Choose a model and plugin. Configure access, approval rules and a model budget.
3. Inspect the result
Review the answer, tool calls and cost. Test interruptions before relying on repeated runs.
Six things oz0 takes care of.
Jobs
Start work, follow its progress and inspect the result.
Rules
Choose which actions run, wait for approval or are refused.
Machines
Put each job on a machine with the tools and access it needs.
Costs
Set spending limits and see what each job uses.
Testing
Compare agent changes on your tasks. Check useful results and model costs.
Access
Manage people, permissions, credentials and recovery keys.