Probing for Tools

Add the ability for each monitor to actively probe tools resources and prompts with real inputs and expected outcomes.

Users can configure one or multiple probers per tool with a specific input payload. The system will then run these probes on a schedule and validate the returned output for consistency robustness and reliability over time.

This would allow teams to:

• Detect silent failures where endpoints are up but behavior is broken

• Validate functional correctness not just availability

• Track long term reliability of agent tools and prompts

• Catch regressions after updates

== Update Jan 10 ==

Scope has been changed from Tools, Resources and Prompts to Tools only.