agentmail-to/agentmail-skills

Overview

This skill gives AI agents their own email inboxes using the AgentMail API so agents can send, receive, and manage email programmatically. It supports attachments, threads, drafts for human approval, real-time notifications, and multi-tenant isolation via pods. Use it to build agent-driven outreach, support bots, or any workflow that needs agent-specific mailboxes and auditability.

How this skill works

Initialize the AgentMail client with an API key, then create inboxes on demand (auto-generated or custom addresses). Send and reply to messages with both text and HTML, attach Base64-encoded files, and group messages into threads. Use drafts to pause for human review, pods for tenant isolation, and webhooks or websockets for real-time events.

When to use it

• Creating autonomous agents that need unique, addressable inboxes
• Sending and receiving emails from agents with attachments and HTML content
• Implementing human-in-the-loop workflows using drafts and approvals
• Building multi-tenant SaaS platforms with isolated customer inboxes (pods)
• Triggering real-time agent actions using webhooks or websockets

Best practices

• Always include both text and HTML versions when sending for best deliverability
• Use clientId idempotency keys on create operations to avoid duplicate inboxes
• Store and manage pod scoping to enforce tenant isolation and access controls
• Label and update message labels to track status (e.g., unreplied, replied)
• Encode attachments in Base64 and validate contentType before sending

Example use cases

• An outreach agent that sends personalized campaigns and tracks replies by label
• A customer-support bot that creates drafts for human approval before escalation
• A SaaS product that provisions a pod per customer and creates isolated inboxes
• A monitoring agent that sends PDF reports as attachments to stakeholders
• A conversational agent that receives inbound email and resumes threads programmatically