What it is
Jev is TypeSafe’s judgment model: it doesn’t write text, it decides. Connect this MCP and the AI you already use (Claude, n8n, Cursor, or any other MCP client) gets a judging specialist on its team. Your AI sends Jev a batch of search terms, ads, products, or placements with a question, and gets a clear answer back for every single item.
Unlike our Google, Microsoft, and Meta Ads MCPs, this one doesn’t read your ad accounts. It judges the data your AI gives it.
Why use it instead of just asking your AI
You can paste 50 search terms into a chat and get decent answers. But a real search term report has thousands of rows, and that’s where chat models struggle: they sample, they drift halfway through the list, and long lists get slow and expensive.
With Jev connected, your AI:
Judges up to 1,000 items in one call, no sampling
Applies the same criteria to every item, so row 900 is judged exactly like row 9
Reports how confident it is per item, so you know which calls to double-check yourself
Does it faster than reading everything in the chat
What that looks like in practice
Prompts you can give your AI once it’s connected:
Pull last month’s search terms with the Google Ads MCP, then use TypeSafe Jev to classify each one as brand, competitor, generic, or irrelevant. Build a negatives list from the irrelevant ones.
Score all our RSA headlines from 1 to 5 on how well they match our brand voice: [describe your voice]. Show me everything scoring 2 or lower.
Flag every placement in this list that is probably a children’s app or game.
It works best combined with our ad platform MCPs: pull the data with one, judge it with Jev, in the same conversation.
Setup
Create an account at typesafe.ai and generate an API key.
Claude: go to Settings → Connectors → Add custom connector.
Name: GAQL.app - TypeSafe Jev by TrueClicks.
URL:
https://mcp.gaql.app/mcp/typesafe/{YOUR-TYPESAFE-API-KEY}Other MCP clients: use the endpoint
https://mcp.gaql.app/mcp/typesafewith your TypeSafe API key as a Bearer token in the Authorization header. If your client can’t set headers, use the URL format from step 2.
🔒 Your key is passed straight through to TypeSafe. We never store it, and no GAQL.app sign-in is needed.
If it doesn’t work
Authentication errors almost always mean a GAQL.app token was used. This MCP needs your TypeSafe API key instead.
A question is rejected: your AI handles the question format on its own. If a question is rejected, the error message says what to fix (for example, too many categories in one question). For how to phrase criteria, see the TypeSafe docs.
✅ You’re all set!
Your AI can now classify, score, and flag PPC data at scale instead of eyeballing samples.
💬 Need help?
Reach out via the chat widget. We’re happy to help you get set up.