How do I validate the TypeScript Montage status contract without an agent?
The official TypeScript client can establish a narrow remote-tool baseline without delegating execution to a model. Inspect the discovered status contract and require a real authorized response before exposing broader research tools, leaving writes and unsupported auth paths explicitly outside the diagnostic. Inspect the real input contract before making the direct call, leaving model-driven tool selection out of readiness. This makes account access observable independently of inference and avoids approving every mutation merely because discovery exposes a complete toolkit.
Help me validate the TypeScript Montage status contract without an agent. State that the diagnostic must return the intended Montage workspace from signals_status, not merely create a Client or list tools. Report the generation, checked account context and actual outcome, then complete cleanup. No writes.Define the expected proof
State that the diagnostic must return the intended Montage workspace from signals_status, not merely create a Client or list tools. Keep provider credentials and any paid model execution unrelated to this test.
Inspect advertised input
Within the approved active connection, discover the status definition through the installed API. Use its real input contract and avoid inventing workspace selectors or borrowing parameters from a research tool with a different purpose.
Separate policy from discovery
Select the one required read before exposing tools to an agent. A complete metadata list is not a user approval grant, and the application must not treat every discovered mutation as automatically permitted.
Execute directly and inspect
Call signals_status through the configured Streamable HTTP client without inference. Preserve actual workspace data and any transport or permission failure; a model-generated explanation cannot substitute for an execution result.
Return path-specific readiness
Report the generation, checked account context and actual outcome, then complete cleanup. The result does not certify another SDK line, grant future writes or establish universal availability across the application’s deployment environments.
Keep the boundaries clear.
- A complete metadata list is not a user approval grant, and the application must not treat every discovered mutation as automatically permitted.
- The result does not certify another SDK line, grant future writes or establish universal availability across the application’s deployment environments.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.