Human-in-the-loop workflows are great for forcing approval steps in AI agents. Read on to see how to build one using @triggerdotdev ↓
1
4
19
Replies
Let's create a wait token, do some work, then pause until an approval arrives (from your UI, Slack, email link, or any service that can POST). While tasks are waiting, you don’t pay a thing.
1
0
1
Completing the approval is simple. From your API route, Slack action, or another service, just call complete. Or you can POST JSON to the token.url that we give you. Great for integrating third-party systems without extra glue code.
1
0
0
Some APIs support webhooks that call back when they're completed. Pass the token.url to them and your task will continue automatically. For example, you can run AI models using @replicate and use their webhook param:
1
0
2
Timeouts are built-in. When you create the token you can specify a timeout period – if nobody responds you can handle the unhappy path with normal code.
1
0
0
Idempotency is built-in. If your task retries, you can reuse the same key to avoid duplicating waits or approvals.
1
0
0
Built-in observability. List and retrieve tokens to audit approvals and tag by user/resource – no custom admin UI required.
1
0
0
Check out this example – it creates audio summaries of newspaper articles using a human-in-the-loop workflow built with ReactFlow and Trigger waitpoint tokens. https://t.co/zcSJIfpkaB
0
0
0