Skip to main content

Agents

An agent is any AI system that connects to Permission Slip to request approval before taking actions. Agents authenticate using Ed25519 key pairs and communicate with Permission Slip through a signed API.

Agent Lifecycle

Agents move through three statuses:
StatusDescription
PendingAgent has registered but hasn’t completed verification yet.
RegisteredAgent is verified and can submit approval requests.
DeactivatedAgent has been permanently disabled. All standing approvals are revoked.

How Agents Authenticate

When an agent registers, it provides an Ed25519 public key. Every subsequent API request from the agent is signed with the corresponding private key using the X-Permission-Slip-Signature header. This ensures:
  • Identity — Permission Slip knows which agent is making the request.
  • Integrity — requests can’t be tampered with in transit.
  • Replay protection — timestamps and request IDs prevent replayed requests.

What Agents Can Do

Once registered and verified, an agent can:
  • Request approvals — submit an action for your review before executing it.
  • Cancel requests — withdraw a pending approval request if no longer needed.
  • Verify confirmation codes — exchange an approval’s confirmation code for a single-use execution token.
  • Execute actions — run an approved action using the execution token.
  • Use standing approvals — execute pre-authorized actions without per-request approval.

Managing Agents