Agent Inbox gives each AI agent a persistent inbox, conversation context, connected tools, and policy-aware execution so email can become completed work instead of another integration queue.
Each agent receives an email identity and an inbox designed to preserve conversation state over time rather than treating every message as an isolated event.
The agent evaluates new messages alongside the relevant history, participants, attachments, prior decisions, and unresolved commitments.
Permissions, prohibited actions, approval thresholds, and escalation rules define what the agent may do autonomously and what must be reviewed.
Email-triggered work can connect to business systems, specialist agents, and internal tools so the agent can retrieve information and perform permitted steps.
The agent can prepare or send the next message within its authority, track what remains open, and continue when a reply or approval arrives.
Action records connect decisions with the context, policies, tools, and approvals involved so teams can understand and review agent behavior.