Glossary

Access policy

An access policy is the set of rules for who and what each AI agent may reach, and what it may do there. It is written where a reviewer can read it, and checked on every call the agent makes.

In a run

The policy is checked on the call itself, before it reaches the tool or the next agent. When a call fails the check, the step shows as refused and the record names what was outside permission. Calls to tools and calls to other agents get the same check.

Our word for it

An access policy you can read.

Who writes it

Your team does, with ours, in the first workflow or the evaluation. For each agent the policy says which systems and actions it may reach. In the evaluation, one step is refused on purpose, so your team sees the policy at work. Creating, changing and running a workflow are separate rights, so you can give them to different people.

A policy you can read

A reviewer can see, before the first run, what each agent may reach and do. After the run, the record shows each place the policy said no.

Policy and credential

The policy sets what an agent may do at all. The scoped credential sets what it may do in this job. A call has to pass both.

Where it stops

  • Ares does not write your business rules. You approve them.
  • The record shows what agents did under the policy, not whether the policy was wise.

The sandbox shows a call refused under the policy, with its reason on the record.

Related terms

  • AI agent access controlAI agent access control decides which systems an AI agent may reach and which actions it may take there.
  • Policy as codePolicy as code means writing rules, such as access rules, as text a machine can check, under the same review and change control as software.

Tell us about one workflow.

We will say whether an agent should do it, and what the first step costs.

A senior member of the team replies within two business days.