Test a tool
Runs the saved tool once, for real, against your endpoint, with values you supply, and shows exactly what happened: whether its condition held, the request as sent (secrets masked), the reply, what your transform made of it, what would be promoted, which template paths were empty, and whether a call would run, skip or abort. It places no call. Use test data for tools that book, charge or send messages.
The body is optional: args (what the model would pass), vars, call and analysis. Check status as well as the HTTP code: a tool that failed still comes back as a 200 with the failed run in it. With this section’s example URL the request can’t connect; point the tool at your own endpoint first.
Authorizations
Your workspace secret key, sk_live_ followed by 32 hex characters. Create one in the console under Developers.
Path Parameters
The agent's id (agt_…).
The tool's name on the agent: lowercase letters, digits and underscores.
Response
Your endpoint answered