Overview & result
This local stdio MCP server gives Cursor, Codex and other MCP clients Jev Choice, Score and Noul judgments for step routing, prepared tool calls, patch review and claim verification. The host agent still edits files and runs commands.
From the original post
A local stdio MCP server that gives Cursor, Codex, and other MCP clients typed TypeSafe Jev judgments. Jev returns Choice, Score, and Noul answers; the host agent still edits files and runs commands.
How Jev fits in the loop
- Expose bounded Jev judgment tools through a local MCP server.
- Let the host agent supply state, candidates, patch or evidence to a matching tool.
- Return typed answers and an auto, review or escalate recommendation.
- Keep file edits, commands and final approval in the host agent.
How to reproduce
- Inspect the source implementation at https://github.com/burnigtm/jev-mcp.
- Configure local Jev API credentials and register required tool/model bindings.
- Execute test prompts and observe decision latency and routing outputs.
Why this build matters
Separates fast typed judgments from execution authority inside coding-agent workflows.
Limitations
- State plus questions must fit an estimated 64,000-token budget; shortened state is reported as incomplete coverage.
- Rank accepts at most 5,000 candidates, with at most 250 options per upstream call.
- Jev only judges; the host agent still edits files and runs commands.