GitHub Copilot per-token billing angers developers

Effective June 1st, GitHub Copilot is shifting its billing model to an AI Credits system, moving away from the previous flat monthly subscription. This new approach will charge users based on their token consumption, encompassing input, output, and cached tokens. This change is particularly contentious for developers who heavily utilize Copilot's agentic coding features, as the new per-token billing can lead to unexpectedly high costs. For instance, a $10/month Pro plan now comes with a credit ceiling that can be depleted rapidly, sometimes within a single afternoon of intensive agentic coding. Workflows involving repository-scale refactors or long-running tasks are reportedly costing between $30 and $40 per run, meaning developers engaged in serious agentic work could exceed their monthly Pro budget well before the month ends. GitHub's Chief Product Officer, Mario Rodriguez, reportedly stated that the flat-rate model was "no longer sustainable" due to the high inference costs driven by agentic sessions. The official GitHub announcement regarding these changes received a significant number of downvotes in community discussions, reflecting strong developer dissatisfaction. While code completions remain free, the economic implications for agent-heavy tasks are leading many to consider alternatives like Cursor or self-hosted Codeium, signaling a potential erosion of developer trust.