Can I track the cost of automated Claude Code runs?

Yes, by using the --output-format json or stream-json flags, you can programmatically extract the total_cost_usd field from the agent's final metadata.

Can I resume a headless session later?

Absolutely. You can capture the session_id from the JSON output of a command and use the --resume flag in a subsequent call to maintain context.

How do I handle permissions in headless mode?

You can use flags like --allowedTools to auto-approve specific actions or --permission-mode acceptEdits to allow the agent to modify files without manual confirmation.

Is there an SDK for programmatic access?

Yes, the Claude Agent SDK provides TypeScript and Python libraries to trigger the Claude Code agentic loop directly from your own application code.

What is Claude Code headless mode?

Headless mode is a non-interactive way to run Claude Code using the -p (print) flag, allowing it to function as a command-line pipeline tool rather than an interactive REPL.

Claude Code Headless Automation

Name: Claude Code Headless Automation
Author: AnExiledDev

byAnExiledDev

•

デプロイメントとDevOps

Automates non-interactive Claude Code workflows through CI/CD integration, JSON stream parsing, and programmatic permission management.

This skill provides a comprehensive framework for running Claude Code as a programmable agent within automated environments. It guides users through utilizing the print mode flag for single-shot tasks, parsing newline-delimited JSON streams for real-time progress tracking, and configuring granular tool permissions for secure, unattended execution. Whether you are building custom DevOps pipelines, automated code review systems, or complex multi-turn scripts, this skill ensures Claude Code operates reliably and cost-effectively without requiring manual human intervention.

主な機能

01Non-interactive execution via the -p flag and stdin piping

02Real-time monitoring with stream-json and structured JSON output

03CI/CD pipeline integration patterns for GitHub Actions and local scripts

04Programmatic cost tracking and budget management per invocation

056 GitHub stars

06Granular tool permissions and auto-approval configurations

ユースケース

01Building custom development utilities using the Claude Agent SDK

02Executing large-scale automated refactoring across entire repositories

03Automating code reviews and security audits within CI/CD pipelines

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add anexileddev/codeforge claude-code-headless

For use in Claude.ai and ChatGPT

Download Skill