# Security and compliance (/extras/security-compliance) Related: [Labels and system labels](/extras/labels-and-system-labels.md), [Deliverability fundamentals](/extras/deliverability-fundamentals.md), [Account lifecycle](/extras/account-lifecycle.md) How AgentMail protects your data and your mail: SOC 2 Type I and Type II compliance, the email authentication protocols behind every domain, and the automatic scanning applied to every inbound message. For procurement, legal, or security-review requests, see [Requesting documents](#requesting-documents) at the end of this page. ## SOC 2 compliance [#soc-2-compliance] > AgentMail has achieved **SOC 2 Type I** (July 2025) and **Type II** (Q1 2026) compliance. ### Current status [#current-status] **Completed July 2025** - Controls properly designed and in place **Completed Q1 2026** - Operational effectiveness validated over time | Phase | Period | Status | | ------------------------------ | ------------------- | --------- | | **Type I Preparation** | June 2025 | Completed | | **Type I Assessment** | July 2025 | Completed | | **Type II Observation Period** | Aug 2025 - Dec 2025 | Completed | | **Type II Certification** | Q1 2026 | Completed | ### What is SOC 2? [#what-is-soc-2] **SOC 2** is an attestation standard by **AICPA** (The American Institute of Certified Public Accountants) evaluating controls over: 1. **Security** - Protection against unauthorized access, both physical and logical 2. **Availability** - System accessibility and operational performance as committed 3. **Processing Integrity** - System processing is complete, valid, accurate, timely, and authorized 4. **Confidentiality** - Information designated as confidential is protected 5. **Privacy** - Personal information is collected, used, retained, disclosed, and disposed per privacy commitments There are two report types: * **Type I**: Verifies that security controls are properly **designed** at a point in time. * **Type II**: Validates that controls **operate effectively** over a period (typically 6–12 months). AgentMail's SOC 2 Type I and Type II reports confirm that our security infrastructure is properly designed, implemented, and operates effectively over time. ### Security controls implemented [#security-controls-implemented] The following controls have been audited and verified as part of our SOC 2 Type I & Type II compliance: **Access control** * Role-based access; **least privilege** enforced * **MFA** (Multi-Factor Authentication) for administrative access and sensitive operations * Quarterly access reviews and revocation upon role change **Encryption & key management** * **TLS 1.2+** for all service/API communications * Data at rest encrypted using industry-standard ciphers * Centralized **KMS** (Key Management Service) for key generation, rotation, and revocation * **Encrypted point-in-time backups** with 30-day retention See [Security Overview](https://agentmail.to/security) for more details. **Email authentication & anti-abuse** * **SPF, DKIM, DMARC** configured across all sending domains * Real-time scanning of inbound/outbound messages for malware/phishing * IP-based **rate limiting** and behavioral abuse detection See [Email authentication](#email-authentication-spf-dkim-dmarc) below for technical details. **Monitoring & incident response** * Centralized logging and anomaly detection with alerting * Documented incident response process: detect → triage → contain → eradicate → recover → post-incident review * Responsible disclosure channel for external security researchers **Resilience, backup & recovery** * Daily encrypted backups with **30-day retention** * Regular **restore tests** to validate RTO/RPO targets * Multi-AZ/high-availability architecture for critical components ### SOC 2 control mapping [#soc-2-control-mapping] | Control Area | Implementation | SOC 2 Criteria | | -------------------- | ---------------------------------------------- | -------------- | | Access Control | RBAC, MFA, quarterly reviews | CC6.1–CC6.7 | | Encryption & KMS | TLS 1.2+, at-rest encryption, key rotation | CC6.8–CC6.9 | | Email Authentication | SPF/DKIM/DMARC, anti-abuse filters | CC7.1–CC7.4 | | Threat Monitoring | Centralized logs, alerts, malware scanning | CC7.2–CC7.4 | | Backup & Recovery | Daily backups, 30-day retention, restore tests | CC7.3 | | Incident Response | Runbooks, post-mortems, disclosure program | CC7.4–CC7.5 | | Workforce Security | Security training, NDAs, background checks | CC5.3–CC5.4 | > The above mappings reflect our audited Type I and Type II controls. ### Type II certification [#type-ii-certification] AgentMail completed the **Type II observation period** (August 2025 - December 2025) and received full **SOC 2 Type II certification** in Q1 2026 from an independent CPA firm. What was validated: * **Continuous Operation**: Controls functioned consistently without gaps * **Change Management**: Security maintained through system updates and changes * **Evidence Collection**: Logs, tickets, training records, access reviews * **Incident Handling**: Real-world response to security events SOC 2 Type II certification provides the highest level of assurance that AgentMail's security controls are not only well-designed but also operate effectively over time. ## Email authentication (SPF, DKIM, DMARC) [#email-authentication-spf-dkim-dmarc] When you add a custom domain to AgentMail, we ask you to add several records to your DNS settings. We understand that this can seem daunting, and we want to be completely transparent about what these records are and why they are necessary. In short, by adding these records, you are giving AgentMail **permission** to do two things: 1. **Send emails on your behalf** that are trusted and pass spam filters. 2. **Receive emails for you** so your agents can process them. This process is standard practice for any third-party email service, and it does **not** give us control over your website or any other part of your domain. Let's break down what each piece does. To prevent spam and phishing, the modern email ecosystem relies on these three core technologies. Our goal is to handle all the complexity of these protocols for you. Your DNS records are simply the way you tell the world that you've authorized us to do so. ### SPF: Sender Policy Framework [#spf-sender-policy-framework] * **What it is:** Think of SPF as a public list of all the servers that are allowed to send email for your domain. * **How it works:** You add a `TXT` record to your DNS that lists the approved IP addresses or domains. When an email server receives a message from `you@your-domain.com`, it checks the SPF record for `your-domain.com`. If the server that sent the email is on that list, the check passes. * **Your Record:** ```text TXT | mail.domain.com | v=spf1 include:amazonses.com -all ``` This record tells the world that AgentMail is an authorized sender for the `mail.domain.com` subdomain. AgentMail sends through Amazon SES, which is why the record authorizes `amazonses.com` rather than an AgentMail hostname. The `-all` part suggests that any server *not* on this list should be considered unauthorized. ### DKIM: DomainKeys Identified Mail [#dkim-domainkeys-identified-mail] * **What it is:** DKIM is like a digital signature for your emails. This signature proves two things: that the email actually came from your domain and that its content hasn't been messed with in transit from you to who you are trying to send to. * **How it works:** We generate a unique, secure key for your domain. When we send an email, we "sign" it with this key. The public part of that key is published in your DNS. Receiving servers use this public key to verify the signature. * **Your Records:** ```text TXT | agentmail._domainkey.example.com | v=DKIM1; k=rsa; p=MIIB... ``` AgentMail gives you a custom DKIM selector host and TXT value. Publishing the key directly at that selector gives receiving servers a precise public key to verify signatures from your domain. Legacy orgs should keep existing working DNS records in place. For new domain setup, add the TXT selector records AgentMail returns. ### DMARC: Domain-based Message Authentication, Reporting, and Conformance [#dmarc-domain-based-message-authentication-reporting-and-conformance] * **What it is:** DMARC is the policy that ties SPF and DKIM together. It tells receiving email servers what to do if an email claims to be from you but fails the SPF or DKIM checks (or both). * **How it works:** You publish a `TXT` record that specifies your policy. You can tell servers to `reject` the message, quarantine it (mark as spam), or do nothing. It also allows you to get reports on which emails are passing and failing these checks. We typically tell servers to reject the message as this increases deliverability. * **Your Record:** ```text TXT | _dmarc.domain.com | v=DMARC1; p=reject; rua=mailto:dmarc@agentmail.to ``` This policy tells servers to `reject` any email that fails authentication. The `rua` tag specifies that aggregate reports about these failures should be sent to `dmarc@agentmail.to`, allowing us to monitor your domain's health and deliverability on your behalf. ### Receiving mail: the MX records [#receiving-mail-the-mx-records] Finally, **MX (Mail Exchange)** records tell the internet where to deliver your agents' email. * **What they are:** MX records are the post office address for your domain's email. * **How they work:** When someone sends an email to `your-agent@your-domain.com`, their mail server looks up the MX record for `your-domain.com` to find out where to send it. * **Your Records:** ```text MX | domain.com | 10 inbound-smtp.us-east-1.amazonaws.com MX | mail.domain.com | 10 feedback-smtp.us-east-1.amazonses.com ``` The hostnames are AWS ones because AgentMail's mail infrastructure runs on Amazon SES. The first record directs all incoming mail for your domain to our servers, so we can ingest it and trigger your agents. The second `feedback-smtp` record is specifically for routing automated feedback, like bounce and complaint notifications from other mail servers, which is crucial for maintaining a healthy sender reputation. ## Spam and virus detection [#spam-and-virus-detection] AgentMail automatically scans every inbound message for spam and viruses before it reaches your inbox. This happens transparently; there is nothing you need to configure. ### Virus detection [#virus-detection] Emails that contain viruses or malware are **rejected at the gateway** and are never stored. Your inboxes will never contain a message flagged as infected. This protects your agents from processing potentially dangerous content. ### Spam detection [#spam-detection] Emails identified as spam are still **stored** in your inbox so you never lose a message that might be a false positive. However, they are **excluded from API results by default** to keep your agent's workflow clean. When you call the [List Threads](/core/conversations#list-the-threads-in-an-inbox) or [List Messages](/core/receive#list-all-messages-in-an-inbox) endpoint, spam messages are filtered out unless you explicitly request them. ### Accessing spam threads [#accessing-spam-threads] To include spam in your results, pass the `include_spam` (`includeSpam` in TypeScript) parameter when listing threads. CLI TypeScript Python ```bash # list threads including spam agentmail inboxes:threads list \ --inbox-id \ --include-spam ``` ```typescript const threads = await client.threads.list({ includeSpam: true, }); ``` ```python threads = client.threads.list( include_spam=True, ) ``` Each thread object includes a `spam` label indicating whether it was flagged as spam, so you can handle flagged threads differently in your application logic. ```json title="Example spam thread" { "thread_id": "thread_abc123", "subject": "You have won a prize!", "labels": ["spam"], "from": "suspicious@example.com", "to": ["your-agent@your-domain.com"] } ``` ## Requesting documents [#requesting-documents] Organizations evaluating AgentMail can request compliance and legal documentation: * **SOC 2 reports** — [request SOC 2 documentation](mailto:security@agentmail.to) * **Subprocessors** — the current list is published at [agentmail.to/legal/subprocessors](https://www.agentmail.to/legal/subprocessors) * **Privacy policy** — [agentmail.to/legal/privacy-policy](https://www.agentmail.to/legal/privacy-policy) * **Security overview** — [agentmail.to/security](https://agentmail.to/security)