The short answer
Create one inbox per agent withcommune.inboxes.create(). Each agent owns its address, builds its own reputation, and receives replies at its own webhook.
Why separate inboxes matter
Reputation isolation. One agent’s bad send → high bounce rate → that inbox gets filtered. The other agents are unaffected. Reply routing. When a prospect replies tooutreach@domain.com, that reply routes to the outreach agent’s webhook. If all agents shared one inbox, you’d have to parse who the reply is for.
Per-inbox metrics. You can see exactly which agent or campaign has a deliverability problem.
Pattern 1: One inbox per agent role
Pattern 2: One inbox per campaign
For outbound sales or marketing agents running multiple campaigns simultaneously:Pattern 3: Shared inbox with routing middleware
One inbox, but extraction classifies intent on inbound, middleware routes to the right agent:hello@domain.com) but different agents handling different topics.
Stagger warmup
If you create 20 inboxes on day 1, each starts a warmup from 0. Don’t create all inboxes simultaneously and immediately send from all of them.Monitor all inboxes
Related
Running 100+ Agent Inboxes at Scale
Full architecture guide for managing large numbers of agent inboxes in production.
Multi-Agent Email Routing
Comparison of routing patterns for directing replies to the right agent.
Inboxes
Full inbox API reference for creating, configuring, and monitoring agent inboxes.

