Test Running An Agent

Product Feedback
1

I really enjoy creating agents, but sometimes they’re doing quite sensitive work. For example editing records in HubSpot. Before I set an agent live, I’d love a way to test how it behaves without touching real, important data.

Ideally this would look like a dry run. Either using dummy data, a sandbox style integration, or a mode where I can see what the agent would change without those changes actually being applied.

That way I can sanity check the behaviour, get comfortable with the output, and only then switch it into a live environment.

Just food for thought, but this would make me much more confident using agents for critical workflows.

1 Like
2

Awesome suggestion. This is something we’d really love to do, and it’s something we’re already experimenting with internally.

Being able to clearly see what actions an agent is going to take, and making sure those actions are secure, is arguably one of the most important parts of using agents for sensitive work.

It would be great to sit down with you and chat through what you had in mind, and see how it lines up with what we’re working on already. I’ll message you to try and arrange a time.

1 Like