AI BriefingCost-Efficient Coding Agent ClawCodex Released
·2026.05.26 14:36
Key point
ClawCodex, an open-source coding agent where a cheap model performs the work and a high-performance model only reviews at decisive moments, has been released.
Details
ClawCodex is an open-source Python coding agent that adopts a 'Worker-Reviewer' structure to maximize cost efficiency.
How it works:
- Worker: A cheap model such as Haiku or DeepSeek performs repetitive tasks like reading files, editing, and running tests.
- Reviewer: A high-performance model such as Opus steps in only at moments requiring decisive judgment (before finalizing an interpretation, before declaring a task complete, when stalled, etc.) to provide advice on 'Gaps / Risks / Do-next'.
Key features:
- Flexible model combinations: Through Client-side mode, you can freely configure various model providers, such as mixing a DeepSeek worker with a Claude Opus reviewer.
- Cost monitoring: The status bar shows token usage and USD cost for the worker and reviewer separately in real time, making it easy to manage spending.
- Performance: It demonstrated strong performance on the SWE-bench Verified test, scoring 58.2% using Gemini 2.5 Pro.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Jevper Released: A Jev Classification Wrapper for OpenAI-Compatible Models with Automatic Logprobs Detection
Whiteboard Launches Open-Source VS Code Fork for Human-AI Collaborative Software DesignOpenComputer Hosts 'Ship Video' Agent Using Claude Opus 5.5 to Generate Explainer Videos from Code