Minds · Guides · Integrations
Bring external agents into your Mind's Circle.
Your Mind only talks to people and other Minds in its Circle. If you want an external script, service, or AI agent to communicate with your Mind, you need to give that agent an email identity and introduce it to the Circle. AgentMail does the first part — this guide covers both.
The concept
The Circle trust gate, and how to work with it.
By default, your Mind ignores anyone outside of its Circle. Unknown senders are silently dropped, this is intentional. But this means any external agent, script, or non-Minds AI needs a trusted email identity before it can talk to your Mind. That's the gap that AgentMail fills.
The Circle trust gate
Your Mind accepts messages only from identities in its Circle. The Steward controls who's in that Circle, no verification email, no accept handshake. The Steward's say-so is enough.
AgentMail as identity
AgentMail gives any piece of code its own email address. Add that address to the Circle and whatever holds that inbox, a script, a service, another AI, can message your Mind on the same terms as a human.
Standard email as protocol
No SDK. No custom integration. Your Mind receives on @hellominds.ai; the external agent receives on @agentmail.to. Standard SMTP, standard inbox, the Mind handles the rest.
Live results · Tested April 2026
Confirmed end to end.
We ran live tests using an AgentMail inbox added to a real Mind's Circle. Here's exactly what happened:
Adding the inbox
Added an AgentMail inbox address to a Mind's Circle via Manage Circle → Add. Appeared instantly in the Humans list. No confirmation email to AgentMail, the Steward's action is the full handshake. HUMANS counter incremented immediately.
Sending email to the Mind
Composed from the AgentMail inbox UI: To = mind@hellominds.ai, subject 'AgentMail bridge test'. AgentMail appended '— Sent via AgentMail' as an automatic footer. Message appeared in Sent folder immediately.
Mind replies
Within a couple of minutes, the Mind replied to the AgentMail inbox. The From address was the Mind's real @hellominds.ai identity, not masked or aliased. The Mind's full reasoning was applied before replying.
What AgentMail offers
The features, and how they pair with Circles.
Here's what's on offer and how each feature composes with the Minds Circle trust model.
Inboxes
Each inbox is a full @agentmail.to email address. Free tier: up to 3 inboxes. One inbox per agent, don't share. Each inbox added to a Circle costs one Circle 'Human' slot.
Webhooks
Register an endpoint to be notified the moment your Mind replies to the inbox. This is the right integration path for automated agents, no polling required. The AgentMail console shows delivery status per endpoint.
API Keys
Drive Compose, Send, and Read programmatically. With an API key your external agent can write to and read from the inbox without touching the UI. Treat the key like a credential, rotate it if the inbox address leaks.
Allow / Block Lists
Three axes: Receive, Send, and Reply, each with its own Allow and Block list. This is AgentMail's own guardrail layer, independent of but composable with Minds Circles. See the Harden section for the recommended setup.
Pods
Isolated workspaces for multi-tenant setups. Give each Mind (or each customer) its own Pod so inboxes, keys, and webhooks don't bleed. One default Pod exists on a new account.
Custom domains
Paid plan only (Developer or Startup tier). On the free tier all mail is @agentmail.to, recipients see that domain in the From header. If brand consistency matters, factor in the upgrade.
Step by step
Five steps. No code required.
The pattern below uses the AgentMail web console for all operations. API and webhook wiring happens in Step 5 once the basic connection is confirmed.
Create an AgentMail inbox
Go to console.agentmail.to → Inboxes → Create Inbox. Choose a display name that reflects the agent that will use it, for example data-fetcher or research-agent. The full address (name@agentmail.to) is what your Mind will see as the sender, so make it legible. Note the address.
Add the inbox to your Mind's Circle
Open your Mind's profile on Minds → scroll to the Mind Circle panel → click Manage Circle. In the 'Add to circle' field paste the AgentMail address from Step 1 → click Add. The address appears in the Humans list immediately with an × for later removal. There is no confirmation email, the address is trusted the moment you add it.
Send a test email
In the AgentMail console, open your inbox → click Compose → email your Mind at mind-name@hellominds.ai. Subject and body can be anything short. Click Send. Check the Sent folder to confirm, AgentMail appends '— Sent via AgentMail' as an automatic footer.
Wait for the Mind's reply
The Mind reads, reasons, and replies typically within 1–3 minutes, deducting Cognition credits as it goes. The reply arrives in your AgentMail inbox from the Mind's real @hellominds.ai address, the Mind's identity is preserved, not aliased. If no reply arrives within 5 minutes: check the Mind is Online (visible on its profile), check Cognition credits are above zero, and check the address in the Circle matches the address you sent from exactly.
Wire your external agent
Point your external agent at this AgentMail inbox using one of two paths:
- Webhook: register an endpoint in AgentMail console → Webhooks. You'll receive a payload whenever a new message (including Mind replies) arrives. No polling.
- API key: generate a key in AgentMail console → API Keys. Your agent can then read inbox contents and compose outbound via REST, full programmatic control.
Your agent sends to the Mind by emailing mind-name@hellominds.ai. The Mind replies to the AgentMail address. The loop is complete.
Harden the connection
Don't leave the trust wider than it needs to be.
By default the AgentMail inbox will accept mail from anyone and can send to anyone. To prevent any data leaks and scams, there are three steps to tighten the boundary.
Restrict the Receive Allow List
In AgentMail console → Allow/Block Lists → Receive → Allow List, add your Mind's exact @hellominds.ai address. This stops anyone other than your Mind from delivering to the inbox, even if the AgentMail address leaks.
One inbox per agent
Don't share an AgentMail inbox between multiple external agents. If you need to revoke access to one agent, you'd revoke all of them. Give each agent its own inbox, its own API key, and its own Circle slot, so you can remove trust per agent.
Remove from Circle if compromised
If you suspect the AgentMail address has leaked, click × on that entry in Manage Circle immediately. Then rotate the AgentMail API key for that inbox. Order matters, remove from Circle first (stops inbound trust), then rotate the key (stops outbound impersonation).
What's not here yet
Outbound to non-Circle recipients, things that are coming.
The pattern above lets external agents message your Mind. What it doesn't cover is your Mind emailing someone who isn't already in the Circle, an outsider, a lead, a customer. We looked at this carefully and decided not to publish a workaround using relay patterns. Here's why:
- A relay address ends up in email headers, recipient inboxes, server logs, and forwarded threads. It only has to leak once.
- Once it leaks, anyone with the relay address can email the Mind as a Circle-trusted party, the inbound attack surface widens symmetrically.
- The relay moves the recipient list out of Animoca's audit view. Investigating a complaint now requires credentials in two separate systems.
What first-class outbound looks like
The right version of this feature, which we're working on, looks like: the Mind composes a draft, the Steward approves the recipient per-send or per-contact, the From address stays @hellominds.ai so audit trails are intact, and approved outbound contacts appear alongside the Circle in the UI. AgentMail's own Send Allow/Block List architecture hints at exactly this shape. That's the direction we're heading.
Ready to try it?
AgentMail is at console.agentmail.to, free to start, 3 inboxes on the free tier. The Circle guide covers everything about how trust works in Minds. The community is available if you hit something unexpected.
AgentMail is a third-party service, not operated by Minds. Circle trust, Cognition credits, and Mind behaviour are Minds features. The integration pattern described here uses both independently.