InductiveClaw

Autonomous iterative coding agent — builds until the vibe is right.

InductiveClaw wraps the Claude Agent SDK in a persistent autonomous loop. Unlike Claude Code, which stops after completing one task and waits for the next prompt, InductiveClaw continuously builds, verifies, evaluates, and iterates on a software project until it reaches a quality threshold or exhausts its budget.

Installation

pip install iclaw

For screenshot-based visual evaluation (optional):

pip install iclaw[screenshot]
playwright install chromium

Requires the Claude Code CLI (npm install -g @anthropic-ai/claude-code) because the Agent SDK shells out to claude. Run claude login once if you plan to use a Max/Pro subscription.

Setup

On first run, iclaw auto-detects your Claude Code login. You can also run guided setup:

iclaw --setup

This walks you through configuring providers (currently Anthropic/Claude; OpenAI and Gemini are planned future features).

Anthropic has blocked third-party projects from routing Max/Pro subscription OAuth tokens through unofficial API gateways. Using claude login locally for personal dev is fine, but keep your usage personal and off public-facing services.

Usage

Interactive mode (default)

iclaw                          # REPL in current directory
iclaw -p ./my-project          # REPL in specific directory

Type /help for available commands, /config to change providers, /quit to exit.

Autonomous mode

# Build a game from scratch
iclaw -g "Build a roguelike deckbuilder with pixel art style"

# Continue on existing project
iclaw -p ./my-game -g "Polish visuals, add sound effects"

# Quick prototype with low quality bar
iclaw -g "Simple snake game in the browser" --threshold 5 --max-iterations 10

# Non-visual project
iclaw -g "Build a CLI tool for CSV analysis" --no-screenshot

How It Works

Interactive mode: A conversational REPL (like Claude Code) with multi-turn context. The agent has access to Bash, file tools, and your project directory.

Autonomous mode:

1. You give it a goal
2. It initializes the project, creates a backlog, and builds the first feature
3. Each iteration: orient, plan, build, verify, evaluate, document, continue
4. Stops when quality threshold is met, max iterations reached, or Ctrl+C

Project state persists on disk (BACKLOG.md, EVALUATIONS.md, source files), so you can stop and resume.

CLI Reference

iclaw [-g GOAL] [--setup] [-p DIR] [-m MODEL] [-t N]
      [--max-iterations N] [--eval-frequency N]
      [--no-screenshot] [--port PORT] [-q] [-v]

Architecture

inductiveclaw/
├── __init__.py
├── __main__.py        # CLI routing (interactive vs autonomous)
├── interactive.py     # REPL mode (ClaudeSDKClient)
├── agent.py           # Autonomous iteration loop
├── setup.py           # Guided provider setup
├── tools.py           # Custom MCP tools
├── config.py          # Configuration + usage tracking
├── display.py         # Terminal output (Rich)
├── providers/         # Multi-provider abstraction
│   ├── anthropic.py   # Claude (implemented)
│   ├── openai.py      # Future feature
│   └── gemini.py      # Future feature
└── prompts/           # Prompt templates (.md files)

Landing Page

The web/ submodule contains the project landing page — a Next.js 16 app with scroll-driven animations, an Art Nouveau-inspired philosophy section, and an animated terminal demo. See docs/web.md for architecture details.

Future Features

• OpenAI Codex — Codex app-server and API key integration
• Gemini — Google OAuth and API key integration
• Provider cycling — automatic failover between providers on rate limits

License

MIT