Skip to main content

​
Inbox Settings console: workspace defaults and per-agent overrides

Inbox → Settings is the hub that tiles every inbox-scoped configuration surface your workspace exposes: Macros, AI deflection, the SLA report, Reply approvals, Tags, Dispositions, Routing, Digital queues, Ticket automation, Agent scripts, and auto-close. Open it at /inbox/settings in the dashboard. Writing any tile requires an owner or admin role (Reply approvals also admits supervisor, the role built to gate it); agents on a lesser role see a read-only view. Five of those tiles govern a workspace default that individual agents or individual conversations can override. This guide walks where each default lives, what it defaults to, how to change it over the API, and the pitfalls of the override order — so an operator reverts a default here instead of chasing it across the queue.

​
1. The five workspace defaults

Each tile below is tenant-owned configuration you set per workspace; nothing here is platform-wide.
  • Reply approvals (/inbox/settings/reply-approvals) — the org-wide fallback that decides whether an agent’s outgoing replies queue for supervisor sign-off. The page shows the org default as a read-only badge and one switch per team member: the per-agent override wins over the org default, and the workspace owner is never gated (server refuses the override on the owner row with 422). The per-workspace AI side of this — the co-pilot kill switch and the auto-approve confidence bar — lives under the inbox_copilot settings key on the AI Deflection tile; both keys default on.
  • SLA defaults (/inbox/sla-report) — first-response and resolution targets for conversations that have no SLA policy of their own. Without a policy the workspace reports against synthetic defaults — a 4-hour first-response clock and a 24-hour resolution clock, business hours only — so the dashboard pill works from day one. Creating a policy (POST /inbox/sla/policies) replaces the synthetic default, and marking a policy is_default makes it the fallback for every conversation no scope-specific policy claims.
  • Routing and auto-assign (/inbox/settings/routing) — the ordered rule list that decides who owns a conversation the moment it arrives. One action, assign_round_robin, is the auto-assign path; when a rule’s named pool has no available operator, the fallback is the longest-idle skilled operator the digital queue’s overflow policy picks (Digital queues tile), so an inbound conversation is never stranded unassigned by a stale pool.
  • Dispositions (/inbox/settings/dispositions) — the outcome-label taxonomy agents and the AI assign when closing a conversation. The taxonomy is a workspace default: the agent’s per-conversation pick wins over anything a routing rule proposes, and an empty pick falls back to the taxonomy’s default label if you define one.
  • Saved views and share links — a saved view pinned team-wide (scope: "team" on POST /inbox/views) is a workspace default sidebar for the whole tenant; share links mint a signed token (TTL 1–90 days) that opens the same filter. Both are reversible — un-pin a view or revoke a link — and neither ships data cross-tenant.
The remaining tiles — Macros, Tags, Ticket automation, Agent scripts, AI deflection budget — are workspace configuration without an agent-level override dimension, so the org default IS the whole story there. They are covered tile-by-tile in the Inbox setup guide.

​
2. Set the defaults over the API

Each default is its own endpoint; there is no single consolidated settings document — read and write per area. Reply approvals. Flip the org fallback by writing a per-agent override for every member, or unset an agent to return them to the org fallback:
Co-pilot workspace defaults. The inbox_copilot settings key holds the kill switch and the auto-approve eligibility bar:
SLA defaults. Replace the synthetic 4h/24h defaults with a workspace policy and mark it the fallback:
Auto-assign. A routing rule with a round-robin pool is the opt-in that makes conversations self-assigning; the fallback pool lives on the digital queue the rule targets.

​
3. Read the defaults back before and after a change

Read each area back through its own endpoint — same GET the dashboard’s tile renders:
The revert path for a per-agent override is to put that user’s row back to the org fallback — flip require_reply_approval to the org value, or re-issue the PUT with the reverted boolean; the settings tile re-seeds from the persisted map on reload, so a re-read shows the reverted state immediately.

​
4. Pitfalls — override order and inheritance

  1. Agent > org default. On reply approvals the resolution order is per-agent override → org-wide default → off. Deleting conversations or agents doesn’t clear the map; stale override rows keep gate-enforcing until you unset them.
  2. Conversation SLA > workspace SLA. A scope-specific policy wins over the is_default fallback; the synthetic 4h/24h pair only applies when no policy at all claims the conversation. Existing conversations keep the policy they opened under until the policy’s targets change.
  3. Routing rules run in ascending priority, lowest first. A workspace-default routing rule named “catch-all” must carry a high priority number, or it beats the specific rules and auto-assign becomes a coin flip across the fallback pool.
  4. Drafts stay advisory regardless of the auto-approve threshold. The co-pilot threshold only marks a draft auto_approve_eligible — a routing hint. Every draft still carries require_human_review: true, an escalate/end suggestion is never eligible, and a reply-gated agent’s accepted draft still queues — so raising the threshold never silently ships a reply.
  5. Business hours pause the clock. business_hours_only: true pauses the SLA timers outside your team’s shifts, so a conversation opened Friday evening doesn’t breach before Monday’s agents arrive. Flipping the flag changes future due-at computation; it does not re-write already-elapsed clocks.

​
See also