A tiered multi-agent architecture for OpenAI Codex designed to optimize token usage and costs. It routes tasks between Sol, Terra, and Luna models based on specific benchmarked capabilities, featuring a low-cost orchestrator and specialized subagents.
Sources & confidenceShowHide
- Models
- Strong
The project is explicitly built around and benchmarks the specific models in the GPT-5.6 family, citing official rate cards and benchmark data.
README says "Built for the GPT-5.6 family (Sol / Terra / Luna)" and provides a detailed credit-per-token breakdown for each.
- Evidence
- Strong
Publicly accessible GitHub repository with complete source code, configuration files, and detailed implementation documentation.
- Discovered
- First posted
- Published on X Jul 12, 2026
- Listed
- Added to Dropday 2d ago