The latest articles on DEV Community by Basant Bhat (@basant_bhat_1). https://dev.to/basant_bhat_1 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4016576%2F262dc0b2-e436-480b-ad95-f51e74ae80f7.jpg https://dev.to/basant_bhat_1 en Basant Bhat Sun, 05 Jul 2026 18:40:00 +0000 https://dev.to/basant_bhat_1/why-your-claude-code-bill-grows-quadratically-and-the-cli-i-built-to-see-it-1kg1 https://dev.to/basant_bhat_1/why-your-claude-code-bill-grows-quadratically-and-the-cli-i-built-to-see-it-1kg1 <p>I blew through my weekly Claude Code quota in 2 days. No warning, no breakdown, no idea which sessions did it. So I dug into how Claude Code actually bills tokens - and built a free CLI to make it visible.</p> <h2> The quadratic cost problem </h2> <p>Claude Code re-sends the entire conversation history and all tool schemas on <strong>every single turn</strong>. That means:</p> <ul> <li>Turn 1 costs 1 unit of input tokens</li> <li>Turn 50 costs ~50 units</li> <li>A 100-turn session costs roughly <strong>4x</strong> a 50-turn session, not 2x</li> </ul> <p>Your costs grow quadratically with session length, and nothing in the UI tells you this.</p> <h2> Where the data lives </h2> <p>Claude Code writes every session as JSONL to:</p> <p>~/.claude/projects//.jsonl</p> <p>Each line contains token usage: input, output, cache writes, cache reads. All the data you need is already on your machine - there's just no tool to read it.</p> <h2> Enter agentburn </h2> <p>I built <a href="https://github.com/Saloni0106/agentburn" rel="noopener noreferrer">agentburn</a>, a zero-dependency CLI that parses those files locally:</p> <p>npx agentburn-cli</p> <p>You get a per-session report: token summary with costs, a "burn curve" sparkline showing cumulative growth per turn, the top token sinks (exact tool calls that cost most), and "fat files" that get repeatedly read.</p> <h2> The doctor command </h2> <p>npx agentburn-cli doctor</p> <p>This scans for waste patterns and gives concrete fixes:</p> <ul> <li>A file read 3 times costing 32k tokens? Add it to .claudeignore.</li> <li>77% cache miss rate? Stop using --resume across days - prompt caches expire after ~1 hour.</li> <li>One Bash call eating 87k tokens? Split large operations.</li> </ul> <h2> Practical tips to lower your bill </h2> <ol> <li> <strong>Keep sessions short.</strong> Because of quadratic growth, two 50-turn sessions are much cheaper than one 100-turn session.</li> <li> <strong>Don't let Claude re-read big files.</strong> Repeated reads of the same file are pure waste.</li> <li> <strong>Mind the cache.</strong> Prompt caches expire after about an hour - resuming an old session pays full price for the entire history.</li> <li> <strong>Measure.</strong> You can't optimize what you can't see.</li> </ol> <h2> It's open source </h2> <p>MIT licensed, zero dependencies, 100% local - your transcripts never leave your machine. If it's useful, <a href="https://github.com/Saloni0106/agentburn" rel="noopener noreferrer">a star on GitHub</a> helps others find it.</p> <p>Feedback and PRs welcome - especially for supporting Codex CLI and Gemini CLI transcript formats.</p> claude