jevXagent
Transparent proxy that asks Jev to classify each coding-agent prompt and injects the decision into the agent's context before it generates.
Overview
jevXagent is a transparent proxy for AI coding agents (Claude Code, Codex, Kilo Code and Cline, auto-detected) that routes each prompt through Jev before the main LLM responds. Jev makes quick classification decisions, such as what kind of task this is and whether to search the codebase or generate code, and the proxy injects them into the agent's context as a block it calls 'Line J', so the agent sees what was already decided and spends fewer tokens deliberating.
The README claims large gains from 'production use', for example response time falling from 2,400 ms to 890 ms and output tokens from 240 to 48, but these are self-reported and the method is not detailed in the excerpt. The design fails open: if Jev times out, the agent still works. It supports streaming, multi-turn context and bypasses Jev for images and PDFs, with live console metrics and JSONL statistics.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | What kind of request is this (e.g. DevOps question or bug fix)? | — |
| Choice | Should the agent search the codebase or generate code? | search · generate |
How it works
- Agent sends a prompt through the proxy
- Proxy asks Jev classification questions
- Decisions are injected into context as Line J
- Agent generates with the pre-made decisions
- Metrics logged; on Jev timeout the request passes through
Strengths
- Fail-open design
- Works with several agents without code changes
- Per-request latency and token metrics
Caveats
- Performance claims are self-reported
- Injected decisions could mislead the agent if wrong
- Adds a proxy hop to every request
At a glance
- Type
- Proxy
- Status
- beta
- Language
- Python
- Runtime
- local proxy
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
When Jev meets LLM — Transparent proxy that makes AI coding agents 3× faster and 80% cheaper. Cut response times from 2.4s to 890ms with Line J architecture.
hellogumbo/awesome-jev
Frequently asked questions
What is Line J?
The proxy's name for injecting Jev's decisions into the agent's context before it generates.
What if Jev is slow or down?
The design fails open and the agent works as normal.
Are the speedups verified?
They are reported by the author from their own use; no independent benchmark is cited.
Related projects
OpenClaw plugin where Jev decides whether a group-chat bot should reply and, optionally, which model tier should answer.
PluginbetaTypeScript- jev-router★ 593
Per-turn model routing for Claude Code and Codex: wraps the real CLI and lets Jev send simple turns to fast models and hard ones to strong.
ProxybetaJavaScript TypeScript/Bun launcher that asks Jev which model and reasoning effort each Codex, Claude Code or Pi turn should use.
ProjectbetaTypeScriptGo CLI model router that uses one Jev analysis to recommend a model, harness and effort level, then selects deterministically.
CLIbetaGoClaude Code and Codex plugin where Jev judges each turn for model routing, delegation, evidence search and compaction.
ProjectbetaTypeScriptClaude Code, Claude Desktop and Codex plugin where Jev reads a task and ranks live models and reasoning effort by quality, speed and cost.
PluginbetaTypeScript