Skip to content
Every command tested · Every cost measured

The Missing Manual
for Claude CLI

Stop guessing what Claude sends. Start controlling it.

Automation, CI/CD, cost control, and security — every command tested with real payloads.

For developers already using Claude Code who want to go deeper.

Terminal
$ claude -p "Summarize this file" --max-budget-usd 0.01 --output-format json
{
"result": "The module exports a retry wrapper...",
"total_cost_usd": 0.152,
"session_id": "sess_7f3a91...",
"usage": { "input_tokens": 4821, "output_tokens": 312 }
}
# Budget was $0.01. Actual cost: $0.152. That's 15x over.
19
zero fluff
60
all tested
10
with real fixes
20
real CLI outputs

You use Claude Code every day.
But do you know what it actually sends?

The interactive mode is intuitive. The programmatic CLI is a different beast — JSON envelopes, session management, tool restrictions, cost control, sandboxing. The official docs cover the basics. This site covers everything else.

Real Payloads

Every JSON envelope, every API response, every cost measurement comes from a real claude-opus-4-6 session.

Tested Commands

60 copy-paste recipes with the exact command, expected output, parsing code, and documented gotchas.

Every Cost Measured

Know the true minimum cost of a CLI call ($0.004), why budgets overshoot by 152x, and how to optimize spend.

Copy. Paste. Ship.

60 tested recipes with the exact command, JSON response, parsing code, and documented gotchas.

Browse All Recipes
Newsletter

New recipes every week

CLI updates, cost-saving patterns, and gotchas — before they hit your production pipeline.

Weekly
Free
Unsubscribe anytime
newsletter

Get CLI tips that ship code

One email per week. Recipes, gotchas, and cost tricks.