jev-pilot
github.com/akramovic1/jev-pilot- Category
- Developer Tools
- Rank
- No. 1663Tools index
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Agent Building
- Interfaces
- Agent Skill / Plugin
- GitHub
- 13 stars
- Latest release
- jev-pilot--v0.12.1
- Date
About
jev-pilot is a Claude Code plugin that, before each turn, queries TypeSafe's Jev decision model with typed questions about the prompt and uses the answers to set the reasoning effort, subagent model, and skill for that turn. Easy work is routed to low-effort, cheaper subagents while hard work gets stronger ones, keeping the main model reserved for the actual conversation.
At a glance
jev-pilot uses a decision model to choose reasoning effort, skills and subagent options for Claude Code turns. It is worth trying if your workload mixes routine prompts with harder tasks and you want to measure how much routing helps.
Good fit
Claude Code users who can evaluate routing choices against a representative set of tasks.
Consider something else when
Set effort manually if prompt text cannot go to the routing backend or you need fixed routing for every run.
Start here
Compare a short routine task and a harder diagnostic task with and without routing. Check the resulting effort choices, task quality and usage before enabling it broadly.
Cost and requirements
MIT-licensed. The project documents a supported Claude Code version and backend options including OpenRouter and TypeSafe. Routed calls can incur additional usage.
Limits to know
The routing decision can be wrong, and the turn waits for a response up to a configured timeout. The project documents sending the prompt and recent messages to the decision backend.
Sources checked 2026-11-05. Source-based assessment.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.