Add jev-use to MCP Servers

[shitianfang]

What this adds

One line at the end of the MCP Servers section:

- [jev-use](https://github.com/shitianfang/jev-use) — Claude Code / Codex / Pi plugin and stdio MCP server. Tools: `jev_judge`, `jev_gate`. Hands the steps that need no text output to Jev's judgment model; a typed escalation contract returns the rest to the LLM. Install: `npx -y jev-use install`

It follows the section's em-dash (—) separator and the Tools: / Install: shape used by the neighbouring AccInt and mcp-interactive-terminal entries. Appended at the end of the section, matching where the most recent entries (Synder Importer MCP, AgentIQ / MoltAd) were added.

Disclosure: I am the author of jev-use and am submitting my own project.

Name disambiguation

jev-use is a crowded name, so to be explicit about which project this is:

• This entry is shitianfang/jev-use — published on npm as jev-use.
• It is not vlad-terin/jev-use, an unrelated repository.
• It is not the examples/jev-use directory shipped inside trycua/cua.

About the project

jev-use hands the agent-loop steps needing no text output to Jev, TypeSafe's judgment model.

• jev_judge batches typed bool/choice/score questions about one state into a single call.
• jev_gate is an opt-in PreToolUse gate that can only deny or ask.
• A typed escalation contract (writing, open_ended, oversized, unsure, unreachable) returns every other step to the LLM rather than guessing. An unreachable backend escalates instead of allowing, so a command that cannot be judged is never waved through.
• TypeSafe, OpenRouter and Vercel AI Gateway backends are interchangeable.
• Install: npx -y jev-use install
• Every demo is a re-runnable script in bench/examples/, and the runs where Jev did worse are published in bench/RESULTS.md.
• MIT licensed.

Only README.md is touched, one line added.