Tool failure simulator
Test how an agent handles errors and tool latency.
Checking this session…
Tools your client can call
Define arguments, a response, and optional failure behavior.
Up to 10 tools. Each accepts an object with string, number, integer, or boolean arguments and returns static text. Set delayMs (0–2000) for latency and isError to true to simulate a failed call. No code is executed.
Open an existing MCP server
About Tool failure simulator
Create a mock MCP server whose tools return controlled errors and optional latency.
When to use it
Create a mock MCP server whose tools return controlled errors and optional latency.
tools/call → isError: trueHow to use it
- Configure the endpoint and choose private or public access.
- Create it and send requests from your client.
- Inspect test output, update or delete the resource when finished.
Limits & privacy
Uses an expiring mock resource, private by default, with separate management and access credentials. Public access is optional. Standard resource quotas, ownership, renewal and deletion apply. Private access tokens are sent in headers, never in share URLs. Static text results, isError and 0–2000 ms delay per tool. This models tool-level failures; it does not execute agent or model code.
Use this tool from the API reference or connect through the SDK, CLI, SSH, or MCP. Hosted calls share your plan allowances.