EulerJT/claudex
103 stars · Last commit 2026-08-11
Run GPT Responses inside the Claude Code harness with protocol-correct tools, agents, streaming, and compaction.
README preview
# Claudex
**Claudex is a compatibility and control plane for running GPT Responses inside the
Claude Code harness—without reducing the integration to a model alias and a changed
base URL.**
The goal is not merely to make GPT answer inside Claude Code. It is to preserve GPT's
Responses reasoning semantics while retaining Claude Code's strongest harness
features: tool execution, permissions, hooks, subagents, editor integration, and
parent/child result delivery.
We summarize that engineering target as `max{Claude Code, Codex}`: keep the stronger
mechanism at each layer. This is a design objective, not an unverified benchmark claim.
The current public baseline is `1.0.0`, with a versioned control plane for Claude Code
`2.1.222` and a Rust bridge snapshot. The public repository contains no production
deployment snapshot, private network topology, internal provenance, credential, log,
conversation, or runtime state.
## Why Claudex matters