Improving token efficiency in GitHub Agentic Workflows

Github·May 7, 2026·Submitted by Mads Kristian Nylund

GitHub built an Effective Tokens (ET) metric — weighting output tokens 4x and cache reads 0.1x with a per-model cost multiplier — to normalize cost across its agentic CI workflows, then used daily Auditor and Optimizer agents to prune unused MCP tools and replace MCP calls with deterministic gh CLI calls, cutting costs by up to 62% on some workflows while being transparent about the confounders (workload shifts, quality tradeoffs) that make interpreting agent KPIs genuinely hard.

More from Github

GitHub Copilot: Meet the new coding agent

GitHub introduces an asynchronous coding agent that takes GitHub issues, spins up a secure dev environment via GitHub Actions, writes code, runs tests, and opens draft pull requests — excelling at low-to-medium complexity tasks in well-tested codebases.

github.blog·1y ago·Agent News·developer-toolscoding-agentcopilot