What AgentMail users actually want

21 sources classify email rendering as a critical security risk. HTML emails are untrusted code that can execute XSS attacks, tracking pixels, and phishing payloads when rendered naively. The research explicitly documents a four-layer defense strategy: DOMPurify sanitization, sandboxed iframes with disabled scripts, strict CSP to block network requests, and safe-source whitelisting for images. Users need to observe agent email activity for oversight and debugging, but the dashboard currently lacks safe rendering infrastructure.

17 sources identify persistent memory as a critical agent requirement, and email provides built-in threading and searchable history. However, the data reveals a mismatch: agents need semantic memory and episodic memory retrieval, but email search indexes only support keyword matching by sender, subject, date, and full text. When clients reference past discussions from months ago, agents must either load entire conversation histories into constrained context windows or fail to retrieve relevant context.

Six sources list attachment handling as a requested feature gap, and six more sources describe agents needing to accept and process documents, review forwarded emails, and participate in workflows involving structured data extraction. Email attachments are a fundamental channel for information exchange, yet AgentMail currently lacks native support for parsing PDFs, extracting text from images, or scanning files for malware before agents process them.

15 sources across two themes describe multi-agent systems requiring dynamic inbox creation and deletion when workflows complete, but the platform lacks automated lifecycle management. Users can create inboxes programmatically via API, but they must manually track which inboxes correspond to which workflows, when to delete them, and how to prevent cost accumulation from abandoned inboxes. The Startup plan costs 200 dollars per month for 150 inboxes, creating financial pressure to clean up unused capacity.

Ten sources describe enterprise requirements for white-label platforms, EU region deployment, and OIDC SSO, but eleven sources reveal uncertainty around platform stability and terms of service. The company reserves the right to modify fees without notice, terminate accounts immediately under sole discretion, and withdraw service at any time. All fees are non-refundable, and the service is provided AS-IS with no warranties for accuracy, reliability, or security. These terms shift commercial risk entirely to users relying on AgentMail for production workflows.

Six sources describe agents needing to maintain context across synchronous voice and asynchronous email channels, being CC'd on existing threads, and managing 50-plus parallel exchanges. LiveKit integration enables voice agents to access AgentMail, but the data reveals a gap: conversation context does not automatically flow between channels. When a voice agent takes a call and then follows up via email, the email thread lacks the call transcript, creating discontinuous context that forces agents to re-establish background information.

Six sources list SDK and MCP protocol integration as requested features, and the research describes agents needing to interact with signups, verification, customer support, and inter-agent communication. Framework integration is currently a manual process requiring users to write custom API clients, creating onboarding friction for teams using popular agent frameworks. MCP protocol support would standardize how agents access email capabilities, reducing integration effort and improving interoperability.