Ian Livingstone is co-founder and CEO of Keycard, which builds runtime authorization infrastructure for AI agents. His work focuses on giving agents task-appropriate access while retaining identity, policy and an audit trail.[1][2][4]
Start with An AI Agent Got Its User Banned From Resy — Keycard, JPMorgan Chase & AAuth →
Links
Keycard biography Personal writingCo-founder and CEO · Keycard
Career
Livingstone’s earlier career includes a Principal Engineer role at Salesforce and platform-strategy leadership at Snyk. Keycard’s biography also identifies him as co-host of The Infra Pod. His writing on software organizations argues that teams need authority aligned with their responsibilities, supported by shared knowledge and infrastructure. That concern with useful autonomy now appears in his work on software agents.[1][4]
Contributions
Keycard verifies the agent, its execution environment and the person it acts for, then evaluates a request against policy. Approved requests receive scoped credentials; denied requests do not. This separates the agent’s identity from a copied human credential and makes access a runtime decision. The company also records authorization events so an organization can inspect which agent requested access and on whose behalf.[2]
Ideas
Livingstone distinguishes broad API scopes from a specific agent mission. The panel considers budgets, transaction history and external policy enforcement as ways to bound actions.[3]
Tools, writing & talks
Writing, research & talks
Start here
A discussion of agent missions, permissions, runtime decisions and auditability.
Start with the panel’s concrete distinction between a broad API permission and a task-specific mission. Familiarity with OAuth scopes helps but is not required for the examples.
Panelist[3]Original source