Skip to content

Documentation

Everything you need to craft production-ready prompts with ClaudeForge.

What is ClaudeForge?

ClaudeForge transforms plain English descriptions into production-ready prompts optimized for specific Claude models. Instead of manually crafting complex prompt structures, you describe what you want in natural language and ClaudeForge does the heavy lifting.

How it works

  1. 1Describe your task — Write what you want Claude to do in plain English.
  2. 2Generate — ClaudeForge creates an optimized prompt in your chosen format and model configuration.
  3. 3Copy and use — Copy the generated prompt and paste it into Claude (or use the API) for best results.

Getting Your API Key

ClaudeForge uses the Anthropic API to generate optimized prompts. Follow these steps to get your API key:

  1. 1Visit the Anthropic Console and sign in (or create an account).
  2. 2Navigate to API Keys in the sidebar.
  3. 3Click Create Key, give it a name, and copy the generated key.
  4. 4Paste the key into the ClaudeForge settings panel. It is stored in your browser only and never sent to our servers.

Privacy note: Your API key is stored exclusively in your browser's local storage. It is sent directly to Anthropic's API and is never transmitted to or stored on ClaudeForge servers.

Output Formats

ClaudeForge supports nine output formats. Choose the one that best fits your workflow:

  • XML (Anthropic)

    Official Anthropic prompt structure using XML tags like <role>, <task>, and <rules>. Recommended for production API usage.

  • TOON

    Block-based format using [ROLE], [TASK], [OUTPUT] sections. Clean and easy to scan.

  • Harness Style

    YAML-like structured format with metadata headers. Great for prompt versioning and testing harnesses.

  • Markdown

    Headers and bullets for human readability. Ideal for documentation and sharing prompts with teammates.

  • Plain Text

    Raw, unformatted prompt text. Use when you need a minimal prompt with no special formatting.

  • JSON

    Machine-readable structured prompt. Perfect for programmatic consumption and pipeline integration.

  • YAML Config

    Key-value configuration style. Convenient for config files and infrastructure-as-code workflows.

  • Claude.md

    Claude Code project rules format for CLAUDE.md files. Ideal for Claude Code and repository-level instructions.

  • System + User Split

    Separate system and user message blocks designed for direct API calls with distinct message roles.

Parameter Guide

Fine-tune how ClaudeForge generates your prompt by adjusting these parameters.

Model

Select the Claude model the generated prompt targets.

  • Opus 4.6— Most powerful. Supports extended thinking and adaptive effort. 128K output, up to 1M context.
  • Sonnet 4.6— Best balance of quality and cost. Supports extended thinking. 64K output.
  • Haiku 4.5— Fastest and cheapest. Ideal for quick tasks. 8K output. No extended thinking.

Extended Thinking

When enabled, Claude thinks step-by-step in a private scratchpad before producing its visible response. This often improves the quality of complex, multi-step tasks. Available on Opus and Sonnet only.

Effort Level

Controls how deeply Claude thinks when extended thinking is enabled.

  • low — Minimal thinking, fastest responses.
  • medium — Balanced thinking depth (default).
  • high — Deeper reasoning for harder problems.
  • max — Maximum thinking budget. Opus only.

Max Tokens

Sets the maximum number of tokens Claude can produce in its response. The upper limit depends on the selected model: 128,000 for Opus, 64,000 for Sonnet, and 8,192 for Haiku. A higher value allows longer outputs but may increase cost and latency.

Tips for Great Prompts

The quality of the generated prompt depends on the quality of your description. Here are some best practices:

  • ✓Be specific about the task. Instead of “write code,” say “write a TypeScript function that validates email addresses using a regex.”
  • ✓Define the output format you expect. Mention whether you want JSON, a numbered list, a code block, or prose.
  • ✓Include examples of desired output. Providing one or two concrete examples dramatically improves prompt accuracy.
  • ✓Specify constraints and edge cases. Mention any limits (max length, allowed values) and how the prompt should handle unusual inputs.
  • ✓Mention the target audience. Telling Claude who the end reader is helps it choose the right tone and level of detail.

Open Source

ClaudeForge is open source and part of the Claude Prompt Engineering Guide project.

View on GitHub

Contributions, bug reports, and feature requests are welcome. Check the repository for contribution guidelines and open issues.