Agent Trust by MNKI

Agent Trust is the control plane for AI agents at work. Every agent gets a verifiable identity, every action is checked against policy, and every decision lands in a tamper-evident ledger. This app is the phone companion for the people who run that fleet. APPROVE WITH EVIDENCE When an agent asks to do something that needs a human – a refund above a limit, a deploy, a data export – you get a push notification. Open it and see exactly what was checked: who the agent is, who delegated the authority, which policy rule fired, and what the limits are. Approve or deny with your fingerprint or face unlock. The decision is signed by your device and recorded in the ledger. KNOW EVERY AGENT Browse the agents in your organisation, their owners, capabilities and risk tier, and the credentials they present. See which ones acted in the last 30 days and which are sitting idle. READ THE LEDGER The audit trail is hash-chained and signed. Scroll decisions and events, open any of them for the full evidence list, and verify the chain from the phone. ACT FAST WHEN IT MATTERS Suspend or revoke an agent, or switch an organisation to emergency deny, with a step-up check that protects the action from a stolen phone. BUILT ON STANDARDS Agent Trust builds on OAuth, OIDC, SPIFFE, AuthZEN, MCP and A2A rather than replacing them, and shows you the raw claims behind every check. The verifier and SDKs are open source. WHO IT IS FOR Security, platform and engineering teams that already run AI agents against real systems, and want to approve, audit and constrain what those agents do. You need an Agent Trust organisation to sign in: create a free one at mnki.com or from the app. No purchases in the app. Plans are managed on the web console at mnki.com.