Social Inbox

Open Dashboard → Social Inbox to work through connected-channel comments and direct messages without losing assignment or status when the next sync runs.

Work the queue

Use Open, Later, and Done as the accountable workflow. Narrow the queue by channel, conversation kind, sentiment, assignee, unread state, or search text. Selecting an item opens its complete persisted transcript and the provider's supported actions.

• Reply to a public comment or send a direct message where the provider permits.
• Leave an internal note that is never included in an AI prompt or sent to the social network.
• Assign a teammate, mark the item Done, move it to Later for 24 hours with an automatic audited reopen, or manually reopen a Later/Done item.
• Use approved reply templates for repeatable responses.
• See attachments, reactions, deletions, sync freshness, reconnect state, and the provider's known limitations in context.

Search covers participant names, handles, conversation previews, provider identities, and persisted message bodies.

Action controls are capability-aware. For example, Instagram comment reactions and LinkedIn direct messages are not presented as supported operations.

Permissions and AI drafts

Team settings provide Engage, View only, and No access permissions. View-only members can read the queue but cannot sync, reply, moderate, assign, or change workflow state.

Draft asks the configured model for a possible response. The generated text is placed in the editor for human review and is never sent automatically. Internal notes are excluded from the prompt. Requesting a draft may use a paid model configured by the deployment.

Consent-aware automations

Open Automations from the Social Inbox header to manage deterministic reply and comment-to-DM rules. New rules default to Paused, Dry run, and Always review. Each revision freezes its channel, conversation types, keywords, response template, consent basis, maximum trigger age, hourly and per-person limits, cooldown, and approval policy.

The Runs tab shows every dry-run, pending, succeeded, skipped, rejected, or failed decision. Engage operators can approve/reject pending responses and retry failed provider actions. The Consent tab is an append-only grant/revoke history; a revoke made while a response waits for approval blocks the send.

Private messages cannot use inbound context alone. They require a versioned keyword opt-in or a current explicit consent record. Repeated provider pages, outbound messages, connected-account identities, paused rules, stale triggers, and previous rule/message decisions cannot create reply loops.

Automation templates support only {{name}} and {{message}}. They do not call the AI model. The separate Draft command remains review-only and may use the deployment's configured paid model.

Provider availability

Comments, direct messages, reactions, and moderation are separate provider capabilities. Availability depends on platform review, granted scopes, account type, and the specific connected asset. The interface hides unsupported actions and keeps missing capabilities explicit instead of simulating success.

If a previously available action disappears, check the channel's sync and reconnect state before changing the workflow. See Connect and manage social channels.

OAuth-authorized agents can use the read-only get_inbox MCP tool. It can inspect queues and sync health but cannot send or moderate content.