What it does
You can watch every call on a local dashboard as the agent works, complete with cost. When Jev cannot be reached the step is marked unscored instead of pretending it scored zero.
Agent tooling
jev-scout
A chaperone for agents doing web research. Nothing gets fetched or trusted without Jev weighing in on it first.
Bookmark: jev-scout
Keep this in your collection.
Leave a noteWhat would you try with this?
: jev-scout
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
A developer can use this code to build an assistant that grades web research. The code connects an AI program to a search engine like DuckDuckGo. It asks an outside AI service called Jev to check if each page is trustworthy.
An access key connects the assistant to Jev. Without a key, the tool just looks for simple words to rate a page. The system allows 300 checks per session. If the app reaches this limit, it continues searching without grading the results.