Autonomous Coding Agent Demo is an autonomous coding solution based on the Claude Agent SDK, utilizing a dual‑agent architecture. One initialization agent parses application specifications and automatically generates a detailed list of features (default 200 test cases) along with a project skeleton. The other coding agent then implements these features incrementally across multiple sessions, intelligently managing and persisting development progress via Git commits and a feature_list.json file, with support for pausing and resuming at any time.
Autonomous Coding Agent Demo incorporates a robust defense‑in‑depth security model, employing OS‑level sandboxing, strict file‑system restrictions, and a configurable Bash command allowlist to ensure the safety and controllability of the code‑generation process. Users can highly customize the application specifications and the number of desired features. They may also choose to integrate an N8N webhook for real‑time monitoring of test pass rates during long‑running sessions, enabling automated construction of complete applications from natural‑language requirements.
First, install the latest versions of Claude Code and the Claude Agent SDK:
# Install Claude Code CLI (latest version required)
npm install -g @anthropic-ai/claude-code
# Install Python dependencies
pip install -r requirements.txt
Verify the installation after completion:
claude --version # Should display the latest version
pip show claude-code-sdk # Confirm the SDK is installed
Copy the sample environment file and add authentication information:
cp .env.example .env
Edit the .env file, configuring at least one authentication method:
# Authentication (at least one required)
ANTHROPIC_API_KEY=your-api-key
# or
CLAUDE_CODE_OAUTH_TOKEN=your-oauth-token
# Optional: N8N webhook for progress notifications
# PROGRESS_N8N_WEBHOOK_URL=https://your-n8n-instance.com/webhook/...
https://console.anthropic.comclaude setup-token to generate one.