Agent-to-agent collaboration (A2A) is becoming a real interoperability story. Inbox Check supports both MCP (tool use from a single agent) and A2A (agent-to-agent delegation) so you can plug deliverability into any part of your automation graph.
When to use which
- MCP: your main agent wants to call Inbox Check as a tool inside its own reasoning loop.
- A2A: your orchestrator delegates deliverability to Inbox Check as a specialist agent with its own sub-plan and long-running tasks.
What the A2A surface looks like
- Task intake. Orchestrator posts a task (“test this draft across Gmail personal + Workspace + Outlook enterprise”). Inbox Check returns a task ID.
- Progress events. Agent subscribes to status updates (“queued”, “sending”, “partial”, “done”).
- Structured results. Per-provider placement, folder, auth verdict, screenshot URLs, recommended action.
- Escalation. If a threshold is breached, Inbox Check can initiate a
notifytask back to the orchestrator.
Example: cold-outbound QA pipeline
A sequencer orchestrator delegates “ship this campaign” into three specialist agents:
- Content agent produces variants.
- Inbox Check agent runs placement tests on each variant, returns placement matrix.
- Sequencer agent picks the top-placement variant and schedules send.
The whole loop is automatic, observable, and auditable.
Try the agent surface
The Inbox Check API is available over MCP and A2A. Zero paid tiers required for normal usage.
FAQ
Which A2A transport is supported?
HTTP+SSE today, WebSocket variant in preview. Clients that follow the published A2A spec work out of the box.
Do I need separate credentials for MCP vs A2A?
No — one token works for both, per our OAuth setup. Unauthenticated use is allowed within rolling limits.
Can I make Inbox Check a sub-agent of my own orchestrator?
Yes — A2A is exactly that pattern. We're the specialist, your orchestrator owns the plan.
Related reading