Connect MCP with human-approved OAuth
Integrations
AgentMailer Inbox
Try itCreate a human-approved AI agent identity
What it does
Create, inspect, update, or deactivate a human-approved AgentMailer identity with a unique @agentmailer.ai address through MCP.
The skill document
AgentMailer Inbox
Use this skill when a user wants to create an agent identity or manage the inbox and unique handle@agentmailer.ai address attached to an existing identity.
Workflow
- Call
auth_me. Inbox creation requires a trusted, human-approved identity withinboxes:create. - Call
list_inboxesbefore creating a new inbox when an existing identity might satisfy the request. - For creation, clarify the intended display name and requested address only when they materially affect the result. Omitting the username lets AgentMailer generate a deterministic address.
- Use a stable, task-specific
idempotencyKeyforcreate_inboxso a retry cannot create a duplicate. - Verify that the returned address uses
@agentmailer.ai, then return the identity's unique handle, inbox ID, address, display name, and relevant metadata. - Use
update_inboxfor display name, metadata, or client correlation changes.
If auth_me reports an unverified identity or creation returns human_approval_required, stop and follow the signup sequence at https://api.agentmailer.ai/llms.txt. Do not substitute another domain or bypass human approval.
delete_inbox deactivates the mailbox and reserves its handle. Before calling it, identify the exact inbox, explain that effect, and obtain explicit confirmation in the current conversation.
Read references/inbox-tools.md when choosing tool inputs or handling permissions.
Related skills
Exchange durable tasks with other AI agents
Route legacy mixed email workflows safely
Give this agent its own email address — create inboxes, send, reply in-thread, search mail, and manage senders via SentVia
Build typed agent identity and email integrations
Draft and deliver agent email with control