Spotify Engineering introduced Portal, a tool that reduced Claude Code token usage by 90%, addressing the high costs of AI coding. The tool was detailed in a September 2026 blog post, highlighting the challenge of token consumption in AI coding agents and the need to optimize usage to manage expenses effectively.
The problem arises because much of AI coding involves input/output tasks rather than complex reasoning, which consumes thousands of tokens unnecessarily. Spotify’s approach was to create two modes that route simpler tasks to cheaper models, reserving expensive frontier models for tasks requiring advanced reasoning. This strategy required no additional subscriptions or platform teams, making it a practical solution for developers.
AI coding costs are projected to exceed the average developer’s salary by 2028, with some engineering teams spending between $200 and $2,000 per developer monthly on tokens. Spotify’s Portal addresses this by significantly cutting token consumption, which is a major expense in AI-assisted development. This cost-saving measure is crucial as AI coding tools become more widespread and token usage surges.
Spotify’s blog post on engineering.atspotify.com provides detailed insights into Portal’s design and impact. The company’s approach demonstrates a scalable method to reduce AI coding costs, which could influence how other organizations manage token usage in AI development environments.