Agentboxd: Comprehensive Email Infrastructure for AI Agents

Overview

Agentboxd is a specialized platform designed to provide real email inboxes for AI agents. It enables these digital assistants to sign up for tools, catch verification codes, and reply in email threads. The platform operates on the principle of "real email in one API call," where every agent receives its own dedicated address. A core feature is the built-in AI triage that checks every inbound email for prompt injection before it reaches the agent.

Benefits

Agentboxd offers several key advantages for developers and businesses using AI agents. First, it allows for instant creation of email addresses through a single API call. Using a consistent client ID ensures that if an agent restarts, it retrieves its existing inbox rather than creating a duplicate. Users can also connect custom subdomains, and the platform handles all necessary DNS records like MX, SPF, DKIM, and DMARC. This ensures that sites blocking throwaway addresses do not block main agent inboxes.

The platform includes a powerful AI triage model that scores every inbound message for category, prompt injection risk, phishing, and urgency. It strips quoted history and signatures from the text provided to the agent, ensuring the agent reads only the new sentence. Every inbound message is checked for authentication compliance, and files are stored with metadata and hashes. The system can extract text from various file types and perform OCR on scans and photos.

Security is a major focus. Drafts created by the agent require human approval before sending. Users can set allow and block lists, apply rate limiting, and use an emergency stop switch to halt all sending in a workspace. Every email sent by an agent includes a header indicating it was sent by an AI agent, providing transparency.

Use Cases

Agentboxd is ideal for scenarios where AI agents need to interact with human users via email. Developers can use the REST API or SDKs in TypeScript and Python to integrate the platform into their applications. The platform supports the Model Context Protocol, making it easy to use with tools like Claude Desktop and Cursor without requiring installation or API key management on the client side.

Businesses can use Agentboxd for automated customer support, where agents handle billing inquiries or sales questions. The human-in-the-loop feature ensures that sensitive communications are reviewed before being sent. Temporary inboxes can be used for one-time verification codes that automatically wipe themselves after a set time. The platform also supports event streams and webhooks for real-time monitoring of inboxes and threads.

Pricing

Pricing details are not available in the provided information.

Vibes

Pricing details are not available in the provided information.

Additional Information

Agentboxd is hosted in France by default, with a second EU server for backups. Enterprise customers can opt for dedicated deployments in the EU or US. The platform is GDPR compliant and includes a Data Processing Addendum. Users control data retention periods, and messages labeled keep never expire. Nightly encrypted backups are restored-tested quarterly.

Recent updates include the listing of Agentboxd as an active remote MCP server in the official MCP Registry. Users previously relying on Smithery's MCP proxy must migrate to the official endpoint as Smithery's managed services shut down. Verified AI agents can opt into the public Rollcall directory to be discoverable by handle, featuring domain-verified badges and OASF records.