Yoshi

Yoshi

A local context-pruning proxy for Claude Code and Codex.

Jev judges new context once, above a configurable size gate. Yoshi applies the
validated omissions while keeping the protocol intact, then forwards the
request to Anthropic or OpenAI and streams the response back unchanged.

Yoshi is a proof of concept. The real version is heading into CompozyOS.
CompozyOS is an open-source operating system for AI agents: it runs Claude Code, Codex and other
ACP agents as durable background sessions. There, context pruning becomes automatic for every
session, the session goal and plan feed the judge, and judging runs between turns instead of on
the request path, which removes the latency you will see below. If that is what you came for,
star CompozyOS and follow along.

What it does

Coding agents resend their whole conversation on every request. Most of that
history is completed tool input, tool output and teammate reports that the
model no longer needs. Yoshi sits between the client and the provider and:

- Judges new spans once: completed Bash, Write and Edit arguments, tool results and teammate reports. The default starts above 50,000 estimated input tokens. Previously forwarded content, human instructions, assistant notes, errors and opaque blocks stay protected. Cached omissions are replayed without rejudging.
- Asks Jev, one span per request, with up to eight requests in flight. Each
  request carries the current task, earlier user constraints, action chronology, paired completion results for input spans and a byte-bounded
  set of retained evidence, and asks two questions: would a required fact be
  lost, and would a binding user constraint be lost.
- Applies only validated omissions. A removed span becomes an explicit
  omission marker; tool IDs, names, file…