features

x47 is packed with features designed for prompt engineers, AI developers, and researchers who need to test and compare LLM outputs efficiently.

Multi-Model Testing

Run the same prompt across multiple LLMs simultaneously. Compare how GPT, Claude, Gemini, DeepSeek, and Llama respond to identical inputs.

  • Test prompts against 5+ providers at once
  • Identify which model performs best for specific tasks
  • Understand different model behaviors and strengths
  • Save time compared to testing models individually

Side-by-Side Comparison

View outputs from multiple models in a split view for easy comparison. Quickly identify differences in tone, accuracy, and completeness.

  • Toggle between tabs and split view modes
  • Compare 2-4 outputs at once
  • One-click copy to clipboard
  • Markdown rendering for formatted outputs

Variable Injection

Create reusable prompt templates with dynamic variables. Use $variable_name syntax to define placeholders that are replaced at runtime.

  • Build templates once, reuse with different inputs
  • Auto-detection of variables in prompts
  • Real-time preview of compiled prompts
  • Perfect for A/B testing prompt variations

Prompt Chaining

Chain multiple prompts together for complex multi-step workflows. Each step can build on the context from previous steps.

  • AI-assisted chain generation
  • Sequential execution with context passing
  • Edit and reorder steps easily
  • Export chains as JSON for reuse

Output Evaluation

Automatically evaluate model outputs using AI. Get scores for task success, instruction following, factuality, safety, and clarity.

  • 5 evaluation metrics per output
  • Visual score bars for quick comparison
  • Detailed feedback and comments
  • Powered by GPT-5.1 for consistent evaluation

History & Persistence

All your prompts and outputs are automatically saved. Browse history, search past runs, and export data for analysis.

  • Automatic saving of all runs
  • Search by keyword, model, or content
  • Up to 1000 history entries
  • Export as JSON for backup

Local-First Privacy

Your data stays in your browser. API keys and history are stored in localStorage, never sent to x47 servers.

  • API keys stored locally only
  • No data collection or tracking
  • Works offline after initial load
  • Full control over your data

Keyboard-First Design

Optimized for power users with comprehensive keyboard shortcuts. Execute prompts, navigate tabs, and manage outputs without touching the mouse.

  • Ctrl+Enter to run prompts
  • Tab navigation shortcuts
  • Quick access to all features
  • Customizable keybindings

workspace_features

Multi-Tab Workspace

Create unlimited tabs for different prompts or projects. Each tab maintains its own system prompt, user prompt, selected models, settings, and variable values. Switch between tabs instantly without losing context.

Persistent State

Your workspace is automatically saved to browser storage. Close the browser and come back later - your tabs, prompts, and settings will be exactly as you left them.

Real-Time Preview

See your compiled prompt with variables replaced before executing. This helps catch errors and ensures your dynamic content is formatted correctly.

model_configuration

Fine-tune how models respond with these adjustable parameters:

Creativity

Controls randomness (temperature). Lower values = more focused, higher = more creative.

Response Length

Set output length from short (256 tokens) to very long (4000 tokens).

Output Format

Request free text, bullet points, tables, or JSON structured output.

Repetition Penalty

Reduce repetitive phrasing by penalizing recently used tokens.

Learn more about settings