Agent test tool
Agent Tool Call Tester
Check an agent’s proposed tool call against a JSON Schema and locate invalid arguments before execution.
Tool Area
Everything runs in the browser.
Agents can choose the right tool but still send the wrong arguments. Paste an MCP-style definition with inputSchema or a function definition with parameters, then supply one proposed call. The tester reports schema errors and leaves the action unexecuted.
- The tester validates data, not authorization or side effects.
- Do not paste private trace data or credentials.
- A valid call may still choose the wrong tool for the user's intent.
How to use
- Paste a tool definition with name, description and inputSchema or parameters.
- Paste the proposed arguments as JSON.
- Run the check, fix the call, and try boundary cases.
Common use cases
- Catch missing required fields before a function call.
- Check enums, numeric bounds and extra properties.
- Create a compact regression case from an agent trace.
FAQ
Will this execute the function?
No. It validates only the definition and provided arguments.
Does it accept OpenAI-style function definitions?
Yes, use a function object with parameters, or an MCP-style object with inputSchema.