Portal by Spotify cut my Claude Code token usage by 90% | Spotify Engineering
Sep 3, 2026Dimitri Mazmanov, Principal Product Manager
Summary
Introduced 'Portal by Spotify' with 'AiKA Modes' to route AI workloads to different models based on task complexity. Developed a 'shunt' Claude Code plugin to enforce this routing via PreToolUse hooks and bash scripts, delegating I/O-heavy tasks (bulk file reading, boilerplate code generation) to cheaper models like Gemini 2.5 Flash. Demonstrated a 90% token usage reduction for specific scenarios and highlighted limitations in delegating editing and complex reasoning.
More from the AI-Assisted Developer Productivity Platform trail
- Why Spotify Is Not Using Bayesian A/B Testing | Spotify Engineering
- AI Changed How Spotify Builds. What We Learned (and Fixed) About Quality at Higher Velocity
- When Can LLMs Replace Humans in A/B Tests? | Spotify Engineering
- Coding Is No Longer the Constraint: Scaling Developer Experience to Teams and Agents at Spotify | Spotify Engineering
- Better Experiments with LLM Evals — A funnel, not a fork | Spotify Engineering