The signal happens. Claude
fires the next move.
Automations watch your accounts, campaigns, and inbox for the moments that matter (a funding round, a new VP, a positive reply) and run the response in real time. Build the rule in the Coil dashboard, or describe it to your AI in chat. Approve once. Then you stop checking anything.
Triggers you watch for. Actions you fire.
Any cell with a checkmark is a one-sentence ask. “When X happens, do Y.” The agent wires the rest.
| Trigger \ Action | Webhook | SmartLead / Instantly | CRM | Slack / Discord | Chat ping | Run a recipe |
|---|---|---|---|---|---|---|
| Funding event | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Senior hire detected | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Product launch | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Reply received | ✓ | ✓ | ✓ | ✓ | ✓ | |
| Bounce-rate spike | ✓ | ✓ | ✓ | ✓ | ✓ | |
| Net-new lead in ICP | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Calendly booking | ✓ | ✓ | ✓ | ✓ | ✓ |
How a new automation goes live.
The agent drafts in chat. You approve once. The server enforces the allowlist on every fire. No surprise webhooks to domains you didn't sanction.
Agent drafts
“Whenever a target raises, draft an opener and POST it to n8n.” The agent proposes trigger, action, fields, and domain.
Allowlist check
Coil validates the webhook domain and confirms only allowed fields will leave. Missing keys or unknown domains fail closed.
One-time approval
You see the exact payload shape and destination before “publish.” This sign-off happens once per automation.
Live + audited
Fires automatically when the trigger matches. Every fire records actor, payload, downstream status, retries.
“What if the agent does something dumb?”
The exact question the next four answers exist for.
Can the agent enable an automation without me?
No. First publish always requires a human admin signature. Once live, the agent can pause, resume, or retire it on its own, but the initial sign-off is yours, and the allowlist of fields and domains is locked at publish time.
What if my webhook is down?
Three retries with exponential backoff, then a chat ping with the failed payload. You decide whether to retry manually or pause the automation. Failed fires queue until you act. Nothing gets silently dropped.
How do I see what's fired?
Ask the agent in chat: "what fired today?" Or open the event stream in the dashboard. Every fire records timestamp, trigger, action, payload, downstream status, and retry count.
What signals can Coil actually detect?
Funding rounds, senior hires, product launches, hiring spikes, tech-stack changes, bounce-rate thresholds, reply sentiment, Calendly bookings, net-new ICP matches. Most signals refresh hourly; some (replies, bounces) are real-time.
Outreach that fires when the moment matters.
Set the trigger in the dashboard or describe it in chat. Approve once. Watch the event stream do the rest.