Jev Gates Grok Tool Calls
Mercante describes Jev as a live gate for Grok 4.7 tool calls in Cursor and Grok Build, deciding whether each call is allowed, requires approval, or is skipped.
mercante@merccanteπ
GROK 4.7 IS $2/$6 IN CURSOR. Jev now decides if the model even wakes up same sticker as 4.6. live in Cursor and Grok Build. CursorBench 4.0: 46.3. DeepSWE v1.1: 71.0. Terminal-Bench 4.0: 37.6. the loop: repo state β Jev Noul/Choice/Score β allow / ask / skip β only then grok-4.7 1 β dump the tool call as state, not a prompt essay 2 β Jev Noul: is this user-requested 3 β Jev Score: blast radius 0-3 4 β Jev Choice: allow, ask, or skip 5 β hard rules still veto. model never gets last word 6 β shadow 200 tool calls. log pick vs what 4.7 would have billed 7 β flip live. you still approve the risky
The loop gives Jev repository state and a tool-call dump, checks whether the action was user-requested, and scores its blast radius from 0β3. Hard rules can still veto actions, and users approve risky ones. The author says to shadow 200 tool calls before switching the gate live.
Also filed under Agents & browsers
- Building a Harness with Jev
Building a Harness with Jev Learn all about Jev, a new System One model from TypeSafe AI, and how you can use it in your agent harness. Jev is up to 200xΒ ...
- How To Build A Harness With Jev | A LangChain x TypeSafe Conversation
How To Build A Harness With Jev | A LangChain x TypeSafe Conversation Hear from Sydney Runkle, Product Manager, OSS at LangChain, Allie Laabs, Founding Developer Advocate at TypeSafe AI, andΒ ...
- Open-Weight Jev-Inspired Decision Model
tev1 is an open-weight Jev-inspired decision model fine-tuned from Qwen3.5 4B.
- MCP Proxy for Screening Agent Tools
Agent Chaperone is an MCP proxy that uses Jev to screen agent tool calls and tool results.