Controlled autonomy
Autopilot. With guardrails that stay on.
Choose what agents can do on their own. Set hard limits that apply to every mode.
Mode: First 10, then auto. First 10 approved by a person, then automatic. Hard limits unchanged.
Outbound controls
Demo data- Mode
- First 10, then auto
- Approval
- First 10 approved by a person, then automatic3 of 10 approved so far
- Daily mailbox cap
- 40 per mailbox (example, you set it)Hard limit
- New mailbox ramp-up
- Four weeks to the full capHard limit
- Country restrictions
- Configured. Unknown countries blockedHard limit
- Unsubscribe protection
- OnHard limit
- Suppression list
- Active, checked before each sendHard limit
- Bounce and complaint pause
- Enabled at 3% and 0.1%Hard limit
- AI budget
- £50 / month (example, you set it)Hard limit
- Banned claims
- ActiveHard limit
Approval mode may change. Hard limits never switch off.
Example values are not advice. The right caps and budgets depend on your domain, your audience and the rules where you sell.
Hard limits
Rules in code, not settings.
These checks run before every outside action, in every mode. The code refuses. It does not just warn. Every decision is written to an audit log.
Suppression before every send
Unsubscribes, hard bounces and complaints go on your suppression list straight away. The list is checked again right before each send.
Unsubscribe on every email
A working unsubscribe link, one-click unsubscribe headers (RFC 8058), a line saying where the data came from and, for US recipients, your postal address.
Daily caps and a ramp-up
Each mailbox has a daily cap you set. New mailboxes start at a third of it and reach the full cap over four weeks.
Auto-pause on bounces and complaints
A mailbox pauses itself when bounces pass 3% or complaints pass 0.1%.
AI budgets per workspace
Daily and monthly AI budgets that you set. Work stops at the limit. Admins see the cost of each run.
Country rules that fail closed
Each country has a rule for cold email. A country with no rule is blocked. Workspaces can make the rules stricter.
No cold email through newsletter services
Keys for Resend, SendGrid, Postmark, Mailgun and Amazon SES are for opted-in mail only. A cold sequence on them is refused.
LinkedIn company page posts only
No scraping, no connection requests and no automated messages. Only posts to your own company page through LinkedIn's official API.
Community members are never leads
People found on GitHub, Reddit, Hacker News, Slack or Discord are never stored as contacts and never emailed.
Banned claims and sources
Sentences in emails need a source you can inspect. Sentences without one, or that use a banned claim, are dropped before anything is sent.
Safeguards and the law
Technical safeguards are not legal advice.
The limits on this page are enforced by the product. They lower the risk of a bad send, a damaged domain or a runaway bill.
They do not make a campaign lawful on their own. You are the sender of every email and the publisher of every post. You need a lawful basis to contact each person, and the rules differ by country. Take your own legal advice.
Read the acceptable use policy, the terms and the data processing agreement.
Build the GTM team you don’t have to hire.
Start with your website. Review your plan. Choose your limits. Let your AI team get to work.
Your tools. Your rules. Your growth.