Skip to content
nofeb

What is Autonomy tiers for email agents?

Autonomy tiers are sending policies attached to an agent's inbox, from read-only to fully autonomous, that the email server enforces on every send. Because the policy lives on the server rather than in the prompt or tool list, an agent cannot exceed it by switching interfaces or by being talked into it.

Checked against the engine on .

How does Autonomy tiers for email agents work?

Leaving the send tool out of an agent's prompt is not a control: if its credential can send, a change of interface or a prompt injection can send. A server-side tier makes the decision where mail is submitted, for every surface at once.

The middle tiers are where most production agents live: every send held for a person, or held unless it goes to an approved list of recipients. A held send needs someone with a separate approval permission, which the agent itself must never hold.

How does nofeb use Autonomy tiers for email agents?

  • Every nofeb inbox carries one tier, enforced for REST, the SDKs, MCP, JMAP and SMTP submission alike. read_only refuses sends; send_with_approval holds every send as awaiting_approval and emits approval.requested; autonomous sends at once, subject to sponsor gating for self-signed-up agents.
  • draft_only refuses sends and leaves the agent writing drafts, which a person edits and sends.
  • send_to_allowlist sends at once to recipients on the inbox's allowlist and holds or refuses the rest, as the inbox's policy says. Recipient blocklists apply on every tier.
  • send_when_confident asks outbound safety questions of each draft (commitments, prices, legal or personal data, new recipients, echoed injection, staying on topic) and sends when every answer clears the inbox's threshold; otherwise it holds the send with the failing questions as the reasons.
  • Caps per inbox and per key (sends per hour and day, recipients per day and per message, new recipients per day) sit beside the tier, so even an autonomous inbox has a budget.
  • Approving needs the send:approve scope, which keys from agent self-signup never get.

Frequently asked questions

How do I require human approval before an AI agent sends an email?

Create the inbox with "autonomy": "send_with_approval". Every send from any interface is held, and a person approves or rejects it with a key that has send:approve.

How do I let an agent email known customers but nobody else?

Put the inbox on send_to_allowlist and add the customers' addresses or domains to its recipient allowlist. Sends to them go out at once; anything else is held for a person or refused.

Status of everything on this page

Every capability named above, with its status as reconciled against the engine.

CapabilityWhat it doesStatus
Autonomy tiers enforced by the server Every inbox carries a sending policy, from read-only to autonomous, enforced on the server for every surface that sends, so switching from the API to MCP cannot bypass it. Shipped
Human approval of held sends A held send waits with status awaiting_approval until a human approves or rejects it; both steps emit events. The approval queue lists every held send with its recipients (each marked if new to the conversation), the reasons it was held and who submitted it. Approving needs the separate send:approve scope, which keys from agent self-signup never hold, so an agent cannot approve its own send. Shipped
Sponsor gating on unverified organizations Until a human sponsor verifies, an organization may only reply, and only to addresses that wrote to it in that thread. Shipped
Drafts Agents write drafts that a person edits and sends, over the API, MCP, JMAP, IMAP and the webmail. It is the documented path for draft-only inboxes: sending a draft needs the approval permission agents do not hold. Shipped
Recipient and sender allowlists and blocklists Lists of addresses, domains or wildcards per organization, pod or inbox: who an agent may write to, and whose mail it reads. A blocked recipient is refused on every autonomy tier, a blocked sender's mail is rejected or quarantined, and you choose whether a send outside the allowlist is held for approval or refused. A check call predicts what a send or an inbound message would meet before it happens. Shipped
Decisions on every message A default set of typed decisions on every inbound email (intent, urgency, needs-reply, automated, verification email, phishing, injection risk, requires-human and language) with a probability each, your own decision schemas beside them, human relabels that correct and retrain it, per-question accuracy and calibration stats, and confidence-gated sending for the send_when_confident tier. Decisions run without leaving nofeb by default; a third-party model sees mail only if you opt in. Shipped
Per-agent rate and recipient caps Limits per inbox and per API key on sends per hour and per day, distinct recipients per day, recipients per message and new recipients per day, checked and charged in one step so concurrent sends cannot overshoot. A usage call shows the budget left, and you choose whether an exhausted cap holds a send for approval or refuses it. Shipped