Pay for what your agent ships.

Every plan counts two things, each with its own monthly allowance. A send is one email your agent dispatches through the API. An inbound reply is one message delivered to one of your agents — includes parse to a structured payload, injection scan, sender-reputation lookup, and webhook delivery. On paid plans, an optional classify step extracts intent + entities + urgency via our 2-call classifier (opt-in per agent, no extra charge).

If your agent already runs an LLM on every reply, leave classification off — you don't need the work done twice. If you want fast routing (switch (reply.intent) with no extra LLM call), turn it on and we'll do the extraction once for you.

Sends stop at your plan's monthly cap until the next billing cycle or an upgrade, and the dashboard warns you at 80%. Inbound replies over the cap are still stored, but they are not scanned or sent to your webhook.

All billing runs through Stripe. Monthly plans bill in advance and there is no annual commitment — but if you would rather pay yearly, Pro is $200/yr and Scale is $990/yr, which is two months free.

Your agent's email rides a managed sending layer with a documented multi-vendor failover runbook and a ~4‑hour DNS swap path, so a single provider outage doesn't take your agent offline. The full provider architecture is documented on the architecture page.

Every send runs through a reputation classifier before it leaves our system — rules plus a lightweight LLM call that score complaint risk. Senders whose score spikes get auto-throttled on shared IPs until reputation builds, and suppression runs at send time so known-bad recipients never get a message. No bans, no account suspensions for borderline behavior. The classifier is invisible to your code — it's deliverability infrastructure, not a product axis you have to choose between.

A per-sender cron monitors complaint and bounce rates every 15 minutes. Any sender crossing 0.3% complaint rate gets paused before the upstream provider flags it at 0.5%. Your agent's deliverability isn't affected by someone else's behavior on the network. For teams that want reputation isolated from the shared pool, the Dedicated IP add-on is available on request (Scale). Worth saying plainly: an IP only earns a reputation from sustained volume, so below a few hundred thousand messages a month the warm shared rail delivers better than your own cold IP.