Inbox & escalation

Escalation rules, human takeover and team roles.

Every conversation from every agent lands in the shared Inbox. Agents (the human kind) can watch the AI reply in real time, take over, leave internal notes, set priority and status, and hand back to the AI.

Escalation rules #

Rules run after every AI turn and decide when a human should step in. Configure them per agent under Settings → Escalation.

Rule	Triggers when
`keyword`	The visitor’s message matches any of the configured phrases (case-insensitive).
`sentiment`	Sentiment score drops below a threshold (e.g. frustrated customer).
`fallback_count`	The agent used its fallback answer N times in a row.
`message_count`	The conversation exceeds N visitor messages without resolution.
`tool_called`	The model called `escalate_to_human`.

Each rule can set status and priority, assign a teammate, pause the AI and email someone. An escalation.created webhook fires as well.

Conversation lifecycle #

Status	Meaning
new	Created, no human has looked at it.
open	Being handled (by AI or a person).
pending	Waiting on the visitor.
closed	Resolved. Reopens automatically if the visitor writes again.

Team roles #

Role	Can
Owner	Everything, including billing and deleting the workspace.
Admin	Manage agents, sources, actions, team, API keys and webhooks.
Agent	Work the inbox: reply, take over, resolve. Read-only elsewhere.
Viewer	Read-only access to agents, conversations and analytics.

Invite teammates from Settings → Team. Members can belong to several workspaces and switch between them from the sidebar.