Early access · Not affiliated with Nous ResearchTurn every email into a controlled agent workflow
Create agent email addresses with one API call. Receive signed webhooks the moment mail arrives, let agents read and draft over MCP — and nothing leaves without human approval unless you say so.
Join the waitlist and we'll email you the moment agent email workflows go live.
curl -X POST https://hermesagentmail.com/api/v1/inboxes \
-H "Authorization: Bearer hm_live_..." \
-H "Content-Type: application/json" \
-d '{ "local_part": "support", "agent_name": "Support Agent" }'
# => { "data": { "address": "support@hermesagentmail.com", ... } }from client@example.com
Approval required
Reason: first-time external sender
Policy: Support Agent cannot auto-send to new contacts
{
"event": "email.received",
"message_id": "9f2c1e4a-…",
"thread_id": "b81d33c0-…",
"address": "support@hermesagentmail.com",
"from": "client@example.com",
"subject": "Pricing for 20 agent inboxes",
"snippet": "Hey — we run a fleet of agents and…",
"requires_human_approval": true
}Agent inboxes are only the beginning.
An inbox gets the email into your system. HermesAgentMail handles what happens next: routing, permissions, memory lookup, task creation, approval, sending, and auditability.
From inbound email to approved action.
A predictable pipeline turns raw inbound email into controlled, auditable agent actions.
Claim an agent address
Spin up an address like support@hermesagentmail.com and connect it to an agent workflow.
Receive safely
HermesAgentMail receives the email, verifies the event, stores metadata, deduplicates it, and queues the work.
Classify and route
The message is routed by workspace, address, sender, policy, and agent role.
Search memory and context
The agent can use approved MCP tools to retrieve thread history, workspace context, or Hermes memory.
Draft, task, or escalate
The agent drafts a reply, creates a task, asks another agent, or escalates to a human.
Approve, send, and audit
Policy decides whether the reply sends automatically or waits for human approval — and every step lands in the activity feed and audit log.
A real API. Not a waitlist promise.
Create inboxes, receive signed webhooks, draft replies, and request sends through a policy-gated REST API — or hand your agent the MCP server and let it work the inbox itself.
POST /api/v1/inboxes
Authorization: Bearer hm_live_...
{
"local_part": "support",
"agent_name": "Support Agent",
"approval_mode": "required"
}{
"data": {
"id": "wf_8c21…",
"address": "support@hermesagentmail.com",
"agent": "Support Agent",
"status": "active",
"approval_mode": "required"
}
}# Draft a reply to an inbound message
POST /api/v1/messages/9f2c1e4a/reply
{ "body": "Thanks for reaching out. For 20 agent inboxes..." }
# => { "outbound_message_id": "om_51…", "approval_required": true }
# Request the send — policy decides
POST /api/v1/outbound/om_51/request-send
# => { "status": "approval_required" } (human approves in dashboard)
# => { "status": "sent" } (trusted inboxes auto-send){
"mcpServers": {
"hermesagentmail": {
"url": "https://hermesagentmail.com/api/mcp",
"headers": { "Authorization": "Bearer hm_live_..." }
}
}
}Trust controls from day one.
HermesAgentMail is an email processor for agent workflows. Customers trust the platform to receive, route, and store email for their agents. We design around transparency, access controls, auditability, and clear permissions.
Start small. Scale safely.
Transparent plans with usage limits that protect deliverability, security, and your costs.
Inbound and outbound emails both count toward monthly usage.
Questions, answered honestly.
What HermesAgentMail is, what it is not, and how it protects your data.
Be first to give your agents controlled email.
HermesAgentMail is not open to the public yet. Join the waitlist and we'll email you when workflows and approvals go live.
Give your agents email — and keep control.
Create root-domain agent addresses, route inbound messages into Hermes workflows, require approvals, create tasks, and make every action visible.