Test a tool
Use the dashboard Test modal to run a tool with sample parameters before attaching it to a live assistant. Tests use stored credentials and hit your real API (or MCP server).
Run a test
Section titled “Run a test”- Open Tools → select your tool → Test.
- Fill path, query, header, and body fields as the model would.
- Click Run.
- Review HTTP status, response JSON, latency, and any result_transform preview.
What to verify
Section titled “What to verify”| Check | Pass criteria |
|---|---|
| Auth | 200/201 — not 401/403 |
| Params | URL matches expected path and query |
| Schema | Response shape fits transform items_path |
| Strict mode | Invalid inputs rejected when strict params enabled |
| Rich content | Cards preview for successful JSON |
Test vs live chat
Section titled “Test vs live chat”| Test modal | Assistant preview |
|---|---|
| You supply all params manually | Model extracts params from conversation |
| Full raw response visible | Same execution path as production |
| No LLM involved | End-to-end behavior |
After modal tests pass, test in assistant preview with natural language.
MCP tools
Section titled “MCP tools”MCP tests list discovered tools from the server. Pick a tool name and provide its parameter schema fields.
Troubleshooting failed tests
Section titled “Troubleshooting failed tests”| Error | Next step |
|---|---|
| 401 | Credential type and token |
| 404 | URL and path params |
| 422 | Body schema vs API expectations |
| Timeout | API latency; consider async execution |
See Troubleshooting tools and execution logs.