yoshi
Context-pruning proxy for Claude Code and Codex: Jev judges which history is still needed, measured not claimed. POC here now, heading soon into [Compozy](https://github.com/compozy/compozy).

- GitHub stars
- 24 as of 2026-09-22
- Forks
- 2
- Language
- TypeScript
- License
- MIT
- Created
- 2026-09-18
- Last updated on GitHub
- 2026-09-18
- Added to AionEdge
- 2026-09-19

What Jev does in yoshi
The project describes itself this way: Context-pruning proxy for Claude Code and Codex: Jev judges which history is still needed, measured not claimed. POC here now, heading soon into [Compozy](https://github.com/compozy/compozy). In the Agent tooling category, Jev typically decides which tool, skill or model to use next, whether a step is finished, and whether an action is risky enough to stop.
Verification
Verified. We confirmed this repository and its Jev relevance through the GitHub API and the source lists below. A specific file link has not been recorded yet.
Topics
- bun
- claude-code
- codex
- context-window
- jev
- llm-proxy
- typesafe
How Agent tooling projects use Jev
Agent tooling is where Jev sees the most experimentation. These projects sit inside coding agents and assistants such as Claude Code, Codex and Pi, and hand the many small, repetitive choices an agent makes to a fast decision model, leaving the large model free for reasoning and writing.
Related Jev projects in Agent tooling
- pi-jev-auto-modejomatsu/pi-jev-auto-modeAdds rule checks to Pi commands and file operations, then uses Jev to assess cases that need further judgment.★ 23
TypeScript - jev-guardleepokai/jev-guardTool-call checks for coding Agents that assess operation risk, user intent and prompt-injection signals.★ 26
JavaScript - jev-cli (Nasrallah-AL)Nasrallah-AL/jev-cliThe jevctl command-line tool connects verification, classification and scoring questions to text input and scripts.★ 20
TypeScript - jev-mcp (blakestone-x)blakestone-x/jev-mcpWraps Jev classification, scoring, yes/no judgments and candidate matching as MCP tools.★ 20
Python - snifftestDanRWilloughby/snifftestA Markdown and plain-text prose linter using local rules and optional Jev judgments.★ 29
TypeScript - jev-axishiftynick/jev-axiA CLI for Jev pick, rate, check, rank, triage, and guard tasks, with optional hooks before Agent tool calls.★ 19
TypeScript