Permit

Know who gave an agent permission to act.

Every grant has an author, a date and a reason. Every agent has an owner and a review date. Every version records what changed and who changed it. Governance here is not a report produced at the end of a quarter — it is the record the platform keeps as a side effect of working.

Build an agent See how it works

Demo data

What this agent may do

Three states per capability. The default for everything is never.

Customer record

Read customersAllowed

Create activitiesAllowed

Change the account ownerNever

Delete a customerNeverNot grantable to any agent, on any plan.

Finance ledger

Read invoicesAllowed

Read payment statusAllowed

Agree a payment arrangementNeeds approvalFinance Manager decides.

Create a credit noteNever

Issue a refundNeverNot grantable to any agent, on any plan.

Evidence as a by-product

Because grants, approvals and versions are recorded when they happen, most governance questions are answerable from the record. A governance process that requires assembling evidence afterwards produces evidence assembled afterwards.

Reviews with a name and a date

Each agent's permissions are reviewed by its owner on a schedule, and the review shows which capabilities have gone unused. Access granted for a reason that expired is the ordinary way this goes wrong.

Widening access is an event

Adding a capability creates a version, returns the agent to review and notifies the owner. An agent whose permissions grew one quiet edit at a time is the failure this prevents.

Separation of duties where it matters

Granting a critical capability requires a second person. The author of an agent cannot be the sole approver of its access to something that changes access.

Questions this answers

The things you currently have to ask somebody.

Each of these is a question an owner asks regularly and currently answers from memory, a spreadsheet, or by interrupting whoever is nearest.

Who approved this agent's access to our finance records?

Which agents are overdue a permission review?

What changed about this agent in the last six months?

Can I produce evidence that a human approved this action?

Where this stops

What it does not do.

The platform records decisions; it does not make them, and it cannot tell you whether a grant was wise. It also governs agents on this platform only — it has no view of automation running elsewhere in your business. And it is not a compliance product: it produces evidence that may be useful to one, which is a smaller and more honest claim.

Where this connects.

Permissions Observability

Give AI a job.

Build an agent, choose what it can access, define what it may do, and deploy it to work safely across your business.

Build an agent Build with the API

No code required · Full SDKs and API · Stop any agent instantly · [email protected]