optim-jev plugin slots a small model into Claude Code to cut token usage
Fit-Fail-3369 · reddit · 2026-10-17
A Reddit user open-sourced optim-jev, an MIT-licensed Claude Code plugin that wires a small model ("Jev") into the harness to reduce token maxing. It ships three experimental tools: context compaction (Jev reviews older tool calls/outputs and keeps, truncates, or drops them while protecting recent messages, as a pre-step to Claude Code's built-in compactor), model + effort routing (choosing between Haiku, Sonnet, and Opus based on the task, relevant files, and evolving project rules), and skill selection (sending only a small relevant skill subset instead of the full catalog). Key design work went into capping Jev's own decision context to avoid quality degradation—prioritizing essential task info, trimming lower-priority history first, and caching repeatable decisions. Benchmarking for token usage, decision quality, and cost is still in progress, but the experiments are repeatable via the repo.
More from coding & agent
- Claude Code Pushes Urgent Security Update Across All Installations — umeshonai · 2026-10-23
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23