BuildWithHQ
Canonical knowledge unit · ai-agents

AI Agents Universal Inbox claim loop

AI Agents work is intended to enter through the Universal Inbox, where eligible items are claimed, leased, heartbeated, and released under user/role permissions instead of being handled through a direct ungoverned inbound shortcut.

ID: BWHQ-GOCLAW-002Status: specifiedAuthority: highai-agentsuniversal-inboxclaimlease

AI Agents Universal Inbox claim loop

Canonical boundary

Inbound work should enter AI Agents through the Universal Inbox. The Build Master explicitly calls for retiring legacy direct inbound helpers that could bypass the Inbox and suggestion-batch approval path.

Claim behavior

The worker claim loop is designed around sp_Inbox_ClaimForGoClaw, lease/heartbeat behavior, and assigned-role checks. If an Inbox item carries an assigned DataRole, the AI Agents owning user must belong to that role before claim.

Safety invariant

Claiming an Inbox item grants work ownership, not unrestricted data access. Every downstream record, relation, file, message, or AI retrieval still uses the normal authorization envelope.

Related knowledge

AI Agents governed AI operator modelAI Agents is BuildWithHQ’s governed AI operator pattern: it works from permitted Inbox/context, prepares actions, uses approval and freshness controls, and executes through canonical action services rather than mutating data directly.Security invariantsSecurity is designed as a shared structural layer: server-derived identity, one authorization path, explicit tenant/application/location scope, and separate permissions for read, AI, export, and sensitive disclosure.
Machine-readable source: canonical Markdown. This HTML companion exists for human reading, search indexing, and AI retrieval.