Codebase context benchmarks

Real-world results measuring token reduction and cost savings on Claude 3.5 Sonnet ($3.00 / 1M input tokens).

Repository Language Total Files Raw Tokens Compressed Reduction Cost Saved / Run
Django ✓ Verified Python 2,359 5.6M 0.6M 89.3% $14.87
Spring Framework ✓ Verified Java 9,193 15.0M 5.2M 65.5% $29.54
Next.js ✓ Verified TypeScript 21,985 24.0M 10.2M 57.4% $41.25
Rails ~ Est. Ruby 6,000 9.0M 2.5M 72.0% ~$19.40
Vue.js ✓ Verified JavaScript 518 1.2M 0.5M 59.7% $2.22
NestJS ✓ Verified TypeScript 1,683 1.0M 0.7M 22.8% $0.65
Vite ✓ Verified TypeScript 1,439 0.8M 0.4M 57.2% $1.42
Tokio ✓ Verified Rust 789 1.6M 0.4M 73.8% $3.53
FastAPI ✓ Verified Python 951 1.1M 0.3M 76.3% $2.59

Test your own project's savings

Run our open-source benchmark utility locally to see the exact line count, token usage, and cost reduction for your repository before installing anything.

$ npx mcp-benchmark ./your-project
Get mcp-benchmark on GitHub →