Agents that dothe work,not imagine it.
Pick a trigger, tell Claude what to do, choose where the result lands. Your first working agent in about 90 seconds, no infrastructure. Like a competitor digest in your inbox at 9am, every day.
Trigger from anything that can send a webhook. Deliver to email, webhooks, and whatever sits behind one:
Three parts.
Zero glue code.
Pick how it fires. Tell the model what to do. Choose where the result lands.
Trigger
Manual run, cron schedule, or an authenticated webhook endpoint you POST to from anything.
Intelligence
Claude does the thinking. System prompt + user template with {{trigger.payload.x}} interpolation.
Action
Email, webhook, log. More coming.
A daily competitor digest, in 90 seconds.
Configure once. The cron handles the rest. Inbox at 09:00 every morning. No webhooks to maintain, no infrastructure to babysit.
- Cron precision down to the minute
- Token usage logged per run
- Retries with exponential backoff
// agents/daily-digest.json{"name": "Daily competitor digest","trigger": { "type": "schedule", "cron": "0 9 * * *" },"model": "claude-sonnet-4-6","system": "You write crisp 5-bullet briefings.","user": "Top moves from {{trigger.payload.topic}}.","action": { "type": "email", "to": "you@co.com" }}
What people actually run.
Churn guard
A payment fails, Claude drafts the save email from the customer's history, it goes out in seconds.
Lead triage
Every inbound form gets scored and summarized. You read one paragraph, not the whole submission.
KPI digest
Each morning Claude pulls your numbers and sends the one-paragraph brief your team actually reads.
Support drafts
Incoming tickets get a drafted reply waiting for review. Your team edits instead of writing from zero.