HeartBeatAgents

Try an Agent Without Digging Through Chat History

View as Markdown

Pick an agent, send it a one-off prompt, and watch a real run play out, kept in its own list instead of mixed into your regular conversations.

What it is

Playground lets you pick an agent and send it a prompt to see how it responds, without creating a conversation you have to go find again in your regular chat list. It isn't a simulation: a Playground run is a real conversation, using the same agent run, the same tools, and the same model as any other conversation. It's just kept in its own Run History instead of your main chat list, with a link from there straight into the full chat view if you want to keep going.

Suggested prompts

When you pick an agent, HeartBeatAgents suggests a few prompts to start with, based on which integrations that agent has connected. An agent with Gmail and Google Calendar connected gets different suggestions than one with GitHub and Linear. A handful of generic suggestions round out the list regardless of what the agent is connected to.

Which agents you can test

The agent picker only lists your organization's active agents. A paused or archived agent won't appear here. With exactly one active agent, Playground selects it automatically.

Working with files

If the agent you're testing has a shared folder connected, you can drag a file into the Playground chat the same way you would in a normal conversation.

  • Agents: configuring the agent you're about to test
  • Shared Folders: connecting a folder an agent can read from during a Playground run
  • Live Preview: what an agent can show you beyond a text reply, during a Playground run same as any other