Orchestrate agent capabilities with PAOL
Connect context, capabilities and approval into a coordinated job.
Begin with a bounded job
PAOL coordinates agent capabilities around a goal. Describe the intended result, required context, allowed actions and the point at which human approval is needed. The same goal can involve multiple capabilities and a workflow.
Define the capability contract
| Contract field | Question it answers |
|---|---|
| Inputs and outputs | What does this capability need and return? |
| Side effects | What changes outside the agent’s reasoning? |
| Autonomy ceiling | What may a user authorize it to do? |
| Approval | Which action waits for confirmation? |
| Failure and recovery | What happens if an action stops halfway through? |
Example: a support exception
- Read the customer request and the approved policy from the permitted context space.
- Prepare a proposed response or exception.
- Request the required approval before the consequential action.
- Execute the approved capability and retain the result for the next step.
Where PAOL runs today
PAOL is the Priostack Agent Orchestration Language. Orchestrations are written, quoted and run from the dashboard while you are signed in. There is no public PAOL API callable with an API key or an agent token yet, so do not build a client against the dashboard routes.
A run is charged 1 credit per action its net fires, and the dashboard shows a quote before it starts; the balance must cover the quote. Reviewing a result is free.
Open the dashboard PAOL overviewMachine-readable contracts: MCP tools JSON · OpenAPI JSON. See the contract notes for older examples.
CONTINUE READINGRun your first workflow →