Hey y’all i've been working on a project that patches parts of Claude Code to make it way faster. It dynamically schedules agents and tool calls based on how many resources they use.

I benchmarked 2–8× faster deep-research, 2× faster code execution, and 3.7–5.5× faster agent swarms. The workflow feature is where it really shows since that's the part that fans out the most work.

reply