GitHub Cuts AI Token Costs by Up to 79% With New Automated Workflow Optimization Tools

GitHub slashes AI token costs by up to 79% using two new daily automation tools — a Token Usage Auditor and Token Optimizer — that identify inefficiencies and streamline agentic workflows across production repositories.

• GitHub is systematically optimizing token usage in its agentic workflows by deploying two daily automation tools — a Token Usage Auditor and a Token Optimizer — that identify inefficiencies, flag costly runs, and recommend concrete fixes across production repositories.
• Key optimizations include pruning unused MCP tool registrations, which reduces per-call context size by 8–12 KB, and replacing GitHub MCP data-fetching calls with direct GitHub CLI commands, eliminating unnecessary LLM reasoning steps and cutting costs without sacrificing output quality.
• Early results across five production workflows show significant reductions in Effective Token consumption, with improvements ranging from 19% to 79%, and the team is now expanding focus to system-level and portfolio-level optimization to eliminate redundant reads and consolidate overlapping workflows.