Sender reputation

Sender reputationis the reputation score Mails.ai attaches to an agent, built from real engagement. The current API returns an agent’s own score. It is designed to propagate across the network as the cohort grows — so a sender’s behavior on one customer’s agent is meant to inform trust calculations on every other customer’s agent that has joined.

What feeds it

• Authentication signals. DKIM pass/fail rate, SPF alignment, DMARC policy + reports, BIMI + ARC where present.
• Behavioral signals across the network. How many of our customers have received from this sender? What were the engagement and complaint rates? Did any of them mark the sender as suppressed?
• Injection-score history. A sender that has consistently sent inbound with high injection_scoreagainst an agent takes a reputation hit — designed to count across customer agents as the network grows.
• Domain age + diversity. Brand-new domains start at neutral 0.5 and earn (or lose) reputation through behavior.

Querying it

Sender reputation is attached to every inbound typed event as sender_reputation(0–1). To pull an agent’s own score directly — built from its send, bounce, complaint, and reply counts over the last 30 days — call the reputation API:

const rep = await client.reputation.get(agentId);

if (rep.reputation < 0.4) {
  // Low standing — tighten sending or pause this agent
}

Identity-level revocation

Global suppression is real today: a suppressed address is suppressed across all workspaces, not just the agent that flagged it. Reputation is designed to work the same way as the network grows — a sender that hits the high-risk injection_score band is meant to take a reputation hit that, as more accounts join, propagates to every Mails.ai customer. The intent is a structural defense against repeat-actor abuse — revocation that follows the sender, not per-account bans that get reset on signup.

Read the architecture page for where the reputation graph sits in the inbound pipeline.