Straight answers, in the first sentence

Email and LinkedIn. Email runs on sending domains you attach. A subdomain, DNS records you publish, verified against the vendor, and LinkedIn runs through your own connected account. SMS is the named next channel and is not open. Instagram, phone and paid ads are out.

No, and that is deliberate rather than a gap. Email runs entirely on attached sending domains, which is what lets cold volume spend a dedicated subdomain's reputation instead of the domain your invoices and password resets come from.

Only at the pace a new account can survive. The realistic ceiling is roughly 20 to 25 connection requests a day for a warmed account and five to ten for one under thirty days old. The ramp is enforced by default. Five a day in week one, ten to day thirty, twenty to day sixty, and it is a switch you can turn off per connection.

Yours. Mail leaves from a domain you own and verified, so the reputation you build is yours and so is the damage a bad list does. That is why the ICP gate, the pacing and the automatic ramp re-anchoring on a bounce spike exist.

Yes, and honestly. Warmup-enabled senders exchange AI-written mundane traffic between your own domains under the same ramp ceiling. Real sends, metered, logged, never shown in your inbox. What it cannot do is build reputation at Gmail or Outlook, because that is earned by mail landing in inboxes there and a domain-only stack cannot fake it.

No. There is one preview before the motion starts and then automatic sending in Autopilot, with approval available in Supervised mode. Supervised is available per campaign for the case where somebody genuinely has to read the copy first. An agency client, usually, and those held steps are reviewed inside that campaign rather than in a global queue.

Because it was a hold rather than a check. The pre-send compliance gate was always the thing standing between a generated message and a person; the queue in front of it asked somebody to re-approve a draft the product had already written from evidence it had already gathered. Queues like that get approved in batches within a fortnight, which is a delay wearing a control's clothes.

A hidden platform prompt layer that nobody can edit away. No invented claims, no implied prior contact, no personal data in notifications, no arguing with the gate. Plus the gate itself, which is not a model call. The database refuses to record a send without a gate verdict and an unsubscribe URL, so a bug fails as a constraint violation rather than as a delivered email.

Every agent prompt, every piece of display copy on an agent, the schedule, the model tier, the send policy, the campaign entry rules and the sequence graph. What you cannot change is the hidden layer above your prompt and the gate below it.

Yes, over the Model Context Protocol. A token carries the authorising person's own permissions, re-read on every call so a demotion takes effect immediately. Nothing in the tool set can send a message to a prospect.

It might, and the risk lands on you rather than on us. Any LinkedIn automation breaches their user agreement and there is no compliant API. Anyone telling you otherwise is wrong. What we do about it: the age ramp is on by default, discovery is budgeted per sweep rather than greedy, and every limit is yours to lower.

We serve the US today and say so rather than implying EU readiness. The controller-grade data model is built. Per-field provenance, retention anchored to collection, a per-row gate, two-scope suppression, because it is cheap to keep and a rewrite to retrofit. An EU deployment is a migration and a legal position, and we would rather have that conversation than sell you a checkbox.

Not yet. The programme is underway and the security page says where it stands. If you need a signed report today, we are not the right choice this quarter.

The United States, in us-east-2, chosen explicitly rather than defaulted into. No EU residency, no adequacy decision claimed.

It will sometimes, and the terms say so plainly: these are AI agents, they make mistakes, and the trade is that good most of the time beats heavily restricted. The failures we engineer against are the irreversible ones. Sending to a suppressed address, sending without an unsubscribe link, continuing after somebody said stop, and each of those is a constraint rather than a prompt.

Yes. Contacts export from the grid as CSV in the table's own order, and the export re-imports fully mapped. Manage cancellation from Billing or contact sales. Purchased credits on usage agreements do not expire at month-end; using the workspace requires an active, paid agreement.