Last month I reached $180 on my Claude Code bill. I use Opus daily: refactoring, bug fixes, code review, feature creation. But something felt wrong. Sessions were getting expensive and I didn't know why.
So I did what any developer would do: I created a tool to find out.
## The experiment
I started manually recording which files Claude reads during a typical session. After a week of monitoring, the pattern was clear:
For every 10 files Claude reads, only 6 or 7 really matter. The rest (configs, README files, lock files, type definitions) are loaded in context and never referenced again.
Quick Math:
-
package.json
read "just in case" → 120 tokens, each time - A README for context → 2400 tokens, used once, forgotten
-
tsconfig.json
→ 400 tokens, never needed
At ~$15/M tokens on Opus, between 30% and 50% of my bill was going to irrelevant context.
## The plugin
I created claude-context-optimizer, a Claude Code plugin that silently tracks every file read, edited and searched. No configuration. No configuration. Just install it and forget it.
Connects to Claude Code's tool pipeline (PostToolUse, SessionStart, SessionEnd) and logs:
- Which files were read and how many times.
- Which files were actually edited (high value)
- Which files were read once and never used (waste)
- Estimated token count per file (~4 tokens/line)
##Context heatmap
Run /cco
and see exactly where your tokens went:
Green bars = files that were actually edited or referenced multiple times. Red bars = read once, never used. Is that README.md eating 2400 tokens? Pure waste.
## Efficiency score
Run /context-digest
for a weekly report card:
You earn a grade (S to F) based on four metrics:
- Context accuracy: how many files read were actually useful
- Editing efficiency: ratio of edits and reads
- Search Accuracy: Do your Grep/Glob searches find the correct files?
- Focus Score: Are you re-reading files too often?
Plus, actual cost breakdowns: total spent, wasted and saveable by month.
## Token ROI Report
/cco-report
gives you a complete picture of all tracked sessions:
Trends over time, most wasted files, most useful files, and specific recommendations.
##How it works under the hood
The architecture is simple: no build steps, no dependencies, just Node.js scripts:
Each session gets a JSON file that tracks the reads, edits, and utility of each file. When a session ends, the plugin updates a global pattern database that learns which files are consistently useful or wasted across sessions.
## The features you didn't expect to need
Token budget: /cco-budget set 80000
sets a token limit with real-time warnings at 50/70/85/95%. Knowing you're at 70% makes you think twice before reading another file "just in case."
Git-compatible hints: /cco-git
analyzes your current difference and suggests files you will actually need. Combined with historical patterns, it knows that when you touch auth-service.ts
, you almost always need user.model.ts
too.
Context Templates: /cco-templates creates bugfix
saves a set of files that you always need. Next time /cco-templates will apply the bug fix
loads exactly the correct context.
## Results after 2 weeks
- Waste ratio reduced from ~40% to ~15%
- Average session cost decreased by approximately 25%
- Sessions feel faster: less context means faster responses
- I stopped reading README.md and package.json reflexively
- Created 3 templates that save ~5 minutes of setup per session.
## The meta part
The entire plugin (2500+ lines of code, 8 forward slash commands, SVG visualizations, README) was created in a single Claude Code session using Opus. Claude created the tool that optimizes Claude. Make of that what you will.
##Try it
Install in one command:
git clone https://github.com/egorfedorov/claude-context-optimizer.git ~/.claude/plugins/claude-context-optimizer
No npm installation, no build, no configuration. MIT lice
