<!-- Envolt Mail · https://envoltmail.com/beta · Markdown version of this page for AI agents. Status: in development (https://envoltmail.com/status). -->

Beta program

# A beta you can trust before it touches your mail

We won't open a beta because a date arrived. Envolt Mail opens to outside teams when eight gates pass, each one measured, and this page shows where each stands. Until then you can join the list and shape what we build.

## Who it's for

### Founders with one inbox

Partners, press, invoices and fans all land in the same place, and every reply sounds like you. You want help without handing over your voice.

### Small teams with shared addresses

hello@, support@ and press@ need answers in the right voice, routed to the right person, with a record of who decided what.

### Builders of AI agents

Your agent needs an address, a verifiable identity and an outbox with rules. [See what it gets](https://envoltmail.com/docs/builders).

## How a pilot will run

1. **Shadow mode.** A read-only connection sits beside your existing mail. Envolt reads, sorts and briefs you; it sends nothing, and your mail keeps flowing exactly as before.
2. **Drafts you approve.** Agents propose replies as decision cards. You approve, edit, hold or delegate each one, with a 60-second window to undo. [Try the cards](https://envoltmail.com/tour).
3. **Autonomy, only if earned.** When one narrow kind of reply has a long, measured record of approvals, you're asked whether to let it go out with notice. You can say no, and any serious mistake takes it back automatically.

Each step needs your say-so. Domains, sending identities and anything with money or legal weight stay with you throughout.

## The eight gates

All eight must pass before any outside team joins. They come from our own launch plan, and they don't move to fit a date.

**0 of 8 met** · 2 in progress · 6 not started

1. [Not started] **90 days on real mail, nothing crossing a line** 
Our own team inbox, read-only, with zero policy breaches and zero leaks between personas, counted from the ledger rather than claimed. The clock starts when the first mailbox is connected.
2. [Not started] **Three teams running on it** 
Three separate teams inside our own studio, each with its own rooms, agents and keys, before anyone outside.
3. [In progress] **Decisions and drafts people actually accept** 
Acceptance above thresholds the accountable person sets, measured with the Autonomy Ledger. The ledger runs in staging on real decisions over synthetic mail; agents do not draft yet.
4. [Not started] **A new team set up in under ten minutes** 
From approving the setup card to running in read-only shadow mode. Waiting on domain changes is reported separately.
5. [In progress] **The red team stays green for 30 days** 
Every release already runs the hostile-mail corpus and leak scans, and no release has let a hostile message trigger an action. The 30-day clock counts on real traffic.
6. [Not started] **Every service level within budget for 30 days** 
Delivery of the morning brief, decision latency and integrity checks, measured on real traffic.
7. [Not started] **Trademark filed and counsel review done** 
Including the data-handling terms a pilot team would sign.
8. [Not started] **A production estate of its own, tested from outside** 
A dedicated production account and database no studio-wide key can reach, customer keys in a key management service, no standing vendor access, and an external penetration test with no open high or critical findings.

## The path there

The planned order, without dates. It may change, and this page will change with it.

### Now

- Agent mailboxes on our own domains, receiving mail in shadow, with daily checks of their records
- The inbound pipeline in staging: Airlock, reader, ledger and integrity checks
- The send path in staging: Seal policy, single-use capabilities, sending limits, a kill switch, signed receipts and labeled messages into a sink
- A read-only console preview for the accountable person, on synthetic test mail
- The approval queue and the Autonomy Ledger in staging, on synthetic mail
- Passkey sign-in for the accountable person, in staging

### Next

- Agent drafting into the approval queue
- Starting the read-only pilot on our own team mailbox, which is built and not started
- Accounts for teams with more than one accountable person

### Later

- Agent replies to outside recipients, labeled and sealed
- Earned autonomy on real mail
- Opening early access to outside teams

## How pricing will work

There are no prices yet. The three rules they will follow are on the [pricing page](https://envoltmail.com/pricing).

## What we'll ask of pilot teams

- Thirty minutes a week of honest feedback, especially when a card was wrong or a brief wasted your time.
- A mailbox you're comfortable connecting read-only, after you've read the [trust center](https://envoltmail.com/trust) and [security model](https://envoltmail.com/security).
- Patience with rough edges: what a beta finds is the point of having one.

The steps a pilot team will take, in order, are in the [getting started guide](https://envoltmail.com/start). The evidence behind each gate is on the [proof page](https://envoltmail.com/proof).

**Join the list.** Tell us what your inbox looks like and what you'd hand to an agent first. We read every request, and we'll contact you before anything about your mail changes. [Request early access](https://envoltmail.com/contact#early-access)
