FAQ

It's early access. The control plane, sandbox engine, and agent-facing HTTP/MCP intents are built and live in test mode, but there is no live production deployment for real sending yet. You can run the full demo pipeline against the live sandbox today with no cost — see the demo walkthrough — but do not tell a user their real campaigns will go out.

A demo tenant is structurally incapable of reaching a real vendor adapter — every domain purchase, mailbox, and send is simulated. There is no live paid/real-sending tier today; real vendor adapters are coded against public API docs but deliberately unactivated, gated behind an owner-hands activation checklist.

Not committed to a date publicly. Real sending is gated behind an activation checklist (Stripe live KYC, real vendor signups, a manual end-to-end deliverability smoke test) documented in the repository's ACTIVATION.md. Join the waitlist to be notified.

No, and no cold-email platform can. Gmail and Microsoft make the placement decision using non-public criteria. This platform makes no deliverability, open-rate, or reply-rate promises anywhere, at any tier. See the deliverability guide for what actually determines placement.

No, and treat any claim like that as inaccurate wherever you see it. Warmup here is legitimate reputation-building over a multi-week ramp — an AI-mimicry layer aimed at generating diverse, non-closed-loop engagement, plus vendor-provided base warmup. It's an incremental edge in an ongoing arms race against spam-filter detection, not a way to defeat filtering.

Domain burn at roughly 8-18%/month is treated as a normal, expected operating cost, not a failure. Outreach runs on branded lookalike domains, never the customer's primary domain, so a burned domain is contained and the platform's deliverability control loop can detect and replace it automatically.

No. Content generation is your coding agent's job — you write the offer and sequence content passed to launch_campaign. This platform owns infrastructure, sequencing, isolation, and guardrails. There is no built-in AI copywriter or autonomous "AI SDR" here.

Signing up mints a sandboxed demo tenant that is structurally incapable of reaching a real vendor adapter — enforced at the type level, covered by a test that fails if the guard is ever bypassed. Every domain purchase, mailbox, and send in demo mode runs against a fault-injecting sandbox (simulated rate limits, timeouts, bounces, asynchronous replies) on a virtual clock, so a multi-week warmup ramp resolves in one session. See the CLI demo details.

Add the hosted MCP server (one config line, streamable HTTP, your bearer token) to Claude Code, Codex, or any MCP-capable agent, or call the HTTP intents directly. Full steps: the setup guide; exact tool schemas: the MCP reference.

Every customer gets dedicated domains and mailboxes, never shared with other tenants. Every authenticated call is scoped to exactly one tenant's state at the engine level, not just at the API layer — there is no cross-tenant data access at any layer.

No. The lookalike-domain generator is scoped to the customer's own brand only. A code-enforced validator rejects a well-known-brand denylist and requires the asserted brand to correspond to the customer's own primary domain — a request like brand=Google with a non-Google primary domain is rejected with a 400.

Per-mailbox/month plus a platform fee: free sandboxed demo, Launch $99/mo (5 mailboxes, 2 domains), Growth $299/mo (20 mailboxes, 6 domains), Scale $799/mo (60 mailboxes, 18 domains), and a metered Custom tier. Full breakdown and what's included at every paid tier: Pricing.

No — month-to-month, cancel anytime. These numbers are live in Stripe test mode and still adjustable until activation.

Compliance is built as a structural requirement, not just policy: a required physical postal address and verified sender identity are injected into every message footer, a suppression list and RFC 8058 one-click unsubscribe are enforced server-side, and the customer is designated the sender of record for their own outreach content and targeting. See the Terms of Service and Acceptable Use Policy (both currently drafts pending attorney review).

You are. For any real (non-sandbox) use, the customer is the "sender" and "initiator" under CAN-SPAM and comparable laws; EpiphanyMade provisions and operates the infrastructure as a service provider acting on the customer's instructions.

agent-cold-email is the permanent keyword handle for the repository, npm package, and MCP registry listings, so none of them ever need renaming. A separate display brand is chosen later, at activation, and shown in one place across this site — the keyword handle stays fixed either way.