Microsoft Research and UC Berkeley release TokenEconomizer to cut coding agent costs
Overview
Microsoft Research and UC Berkeley released TokenEconomizer, an open-source plugin that learns from past sessions to reduce coding agent spending.
On 277 SWE-Bench tasks, MarkTechPost reports, it lowered mean cost by 18.4% to 30.9% across five agent-model settings, with Pass@1 rising in four of them.
The plugin works with Claude Code, Codex, GitHub Copilot CLI and Pi. It does not modify model weights or agent code, and its learned data stays local in ~/.tokeneconomizer.
Written by AI from the articles below · updated Oct 30, 3:49 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- MarkTechPostNewsPickMicrosoft's TokenEconomizer plugin cuts coding agent costs by up to 30.9%
AIMicrosoft Research and UC Berkeley researchers released TokenEconomizer, an open-source plugin that learns from past sessions to reduce coding agent spending. On 277 SWE-Bench tasks it lowered mean cost by 18.4% to 30.9% across five agent-model settings, with Pass@1 rising in four of them. The plugin works with Claude Code, Codex, GitHub Copilot CLI and Pi, and it does not modify model weights or agent code.
Heat Trend
Not enough continuous observations to show a trend yet.