Scoped credentials
API keys are bound to one agent and its permitted actions, and can be revoked.
Controls you can verify.
AgentLedger is designed to reduce the blast radius of autonomous actions while preserving evidence for review.
Pre-execution policy
AgentLedger returns allow, block, or require_approval before your executor performs the proposed action.
Human approval
Sensitive actions can pause until an authorized workspace member approves or denies them.
Signed outcomes
Completion events use HMAC-SHA256 signatures, timestamps, delivery identifiers, and bounded retries.
Audit evidence
Requests, decisions, policies, approvers, timestamps, and execution outcomes remain connected.
Secrets stay server-side
Model and executor credentials remain in your application. AgentLedger keys and signing secrets should never be exposed in client code.
Operational security
Traffic to AgentLedger uses HTTPS. Workspaces and records are associated with an organization, privileged service operations run on the server, and webhook destinations must be public HTTPS endpoints. We monitor and maintain the service with a least-privilege approach to credentials.
Your responsibilities
Store API keys and webhook secrets in a server-side secret manager, grant only the actions an agent needs, review policies and approver access regularly, rotate exposed credentials, validate webhook signatures against the raw request body, and keep your connected executors idempotent.
Responsible disclosure
If you believe you found a vulnerability, email security@agentledgersecurity.com with reproduction steps and potential impact. Please do not access or modify data that is not yours, disrupt the service, or publicly disclose an unresolved issue.
Security posture, stated plainly
AgentLedger does not claim a certification or compliance attestation that has not been independently completed. Contact us for current security information needed for your review.