Give agents a job
to do together.

Where PAOL fits

PAOL is the Priostack Agent Orchestration Language. It brings coordinated agent work into the platform’s execution model. Use it when a goal involves several capabilities rather than one isolated response.

Shared context can inform the work; workflow models can structure repeatable steps. The application declares what it can do, and the permission model determines what may actually run.

A concrete example: investigate a service issue

1. A support agent reads the reported problem and the context it is allowed to use.
2. A diagnostic capability checks the relevant system.
3. A decision routes a routine fix or a request for human review.
4. An approved action is passed to the appropriate service.
5. The result is made available for the next step.

Capability comes before autonomy

Applications describe capabilities, side effects and maximum autonomy. A user’s grant can narrow that ceiling. Financial commitments and account-security actions are capped at confirmation in Priostack’s published capability model.

See the quote before a run

A run costs 1 credit per action its net fires: a task, a decision, a case item, an integration step or an agent memory call. Flows, events and gateway branches are free. Your account shows the quote, a minimum and a maximum, before the run starts; the balance must cover it, and reviewing the result is free.

Use the actual run quote for budgeting. A shared-memory estimate is not a PAOL quote.

Start from a real workflow

Describe the desired outcome, the capabilities you have, and the decisions that need confirmation. Write the definition in your account, where each run shows its quote first. There is no public PAOL API yet; contact the team if you want to talk a design through.