eagle-lang 0.1.1

An AI-native programming language and interpreter.

Eagle: The natural language platform for orchestrating custom, evolving AI agents

Natural language platform with advanced scripting capabilities

Eagle is a natural language platform where you write instructions in plain English instead of code. Write .caw files with plain English, and Eagle's AI interprets and executes them using a comprehensive and customizable set of tools. Rather than generating code for you to run, Eagle directly executes your intent. Eagle can edit itself, call itself, and write its own caw files.

What Makes Eagle Different

• 🗣️ Natural Language Interface: Write instructions in plain English instead of learning syntax
• 🗣️ Multi-step Orchestrator: The Eagle interpreter is a multistep tool calling AI orchestrator
• 🚀 Multi-Provider AI: Works with OpenAI, Claude, Gemini, or OpenRouter
• 🔄 Self-Modifying Architecture: Eagle can create tools for itself, modify its own configuration, and generate new behavior rules
• ⚙️ Easy tool creation: Tools are python scripts that are easy to spin up. Need memory, access to email, or google drive? ask eagle to make a tool.
• ⚙️ Zero-Config Tool Sharing: Tools are self-contained that work across projects and teams
• 🔗 Workflow Composition: Eagle understands how tools work together and suggests intelligent multi-step workflows
• 🔄 Recursive Intelligence: Eagle can call itself to break down complex tasks, delegate subtasks, and coordinate multi-step solutions
• 💬 Interactive REPL: Type commands directly in natural language for immediate execution and feedback
• ⚡ Smooth Setup: Guided configuration with automatic .env creation and API key help

Installation

pip install eagle-lang

Quick Start

Option 1: Script Files

1. Initialize Eagle in your project:

eagle init

2. Create a .caw file with your instructions:

# my_task.caw
Help me write 3 tweets about the launch of a new natural language platform called Eagle.
Focus on its ability to use plain English and orchestrate AI agents. Use my_tone.txt to define my tone.

Option 2: Interactive Mode

# Install and initialize
pip install eagle-lang
eagle init

# Start interactive mode
eagle

Then type commands directly:

eagle> create a python script that sorts a list of numbers
eagle> test it with the numbers [3, 1, 4, 1, 5, 9]
eagle> save the script as sort_demo.py

3. Run Eagle:

eagle my_task.caw

Eagle will interpret your instructions and execute them using available tools.

Configuration

Interactive Setup

Run eagle init for guided setup, or use eagle init -g for global configuration.

Manual Configuration

Eagle looks for configuration in this order:

1. Project: .eagle/eagle_config.json
2. User home: ~/.eagle/eagle_config.json

Example configuration:

{
  "provider": "openai",
  "model": "gpt-4o",
  "rules": ["eagle_rules.md"],
  "tools": {
    "allowed": ["print", "read", "search", "ask_permission"],
    "require_permission": [
      "write",
      "shell",
      "web",
      "git",
      "make_tool",
      "edit_config"
    ]
  },
  "max_tokens": 4000
}

🔄 Auto-Updates: When you create tools with make_tool, Eagle automatically updates this config

API Keys

Set your API key in .eagle/.env:

OPENAI_API_KEY=your-api-key-here
ANTHROPIC_API_KEY=your-claude-key-here
GOOGLE_API_KEY=your-gemini-key-here
OPENROUTER_API_KEY=your-openrouter-key-here

Usage

Basic Commands

# Interactive mode
eagle                         # Start interactive REPL

# Initialize configuration
eagle init                    # Project configuration
eagle init -g                 # Global configuration

# Run .caw files
eagle my_task.caw             # Simple syntax (recommended)
eagle run my_task.caw         # Explicit syntax
eagle task.caw --provider claude --model claude-3-sonnet
eagle task.caw --rules custom_rules.md

# Explore capabilities
eagle capabilities            # Show current workflows and tools
eagle capabilities --detailed # Show technical tool information

Interactive Mode

Eagle includes a powerful REPL (Read-Eval-Print Loop) for immediate interaction:

$ eagle
🦅 Eagle Interactive Mode
Type your instructions in plain English and press Enter.
Commands: .exit (quit), .help (show help), .config (show config)
============================================================
Ready! Using openai:gpt-4o

eagle> create a simple hello world python script
--- Eagle is thinking... ---
[Eagle creates the script using available tools]

eagle> run the script
--- Eagle is thinking... ---
[Eagle executes the script and shows output]

eagle> .exit
Goodbye! 🦅

Interactive Commands:

• .exit / .quit - Exit Eagle
• .help - Show available commands
• .config - Show current configuration
• .capabilities - Show available tools and workflows

Command Options

• --provider: AI provider (openai, claude, gemini, openrouter)
• --model: Specific model to use
• --rules: Custom rules files (space-separated)

Built-in Tools

Eagle comes with a comprehensive toolkit for software development:

File Operations

• write: Create/modify files with intelligent directory creation
• read: Read files with line ranges and multi-file support
• search: Search files by name patterns or content with regex

System Operations

• shell: Execute shell commands with safety guardrails
• git: Full git workflow (status, commit, push, branch management)

Communication & I/O

• print: Styled terminal output with formatting options
• ask_permission: Interactive user input and confirmations
• web: HTTP requests, API calls, and web scraping

Meta-Programming Tools

• make_tool: Describe a tool in English, Eagle generates the code automatically
• make_rule: Generate custom behavior rules for Eagle
• edit_config: Modify Eagle configuration programmatically
• call_eagle: Recursive Eagle calls for complex task delegation

Tip: Use eagle capabilities to see what's possible with your current setup

Examples

File Management

# organize.caw
Look through all the Python files in this directory and organize them by creating
subdirectories based on their purpose (models, views, utilities, etc.).
Move the files to appropriate directories and create an index.md file
explaining the new structure.

Development Workflow

# deploy.caw
Check the git status, run the tests, and if they pass, commit all changes
with an appropriate message and push to the main branch. If tests fail,
show me what failed and ask for permission before proceeding.

Data Analysis

# analyze.caw
Read the sales_data.csv file, analyze the trends, and create a summary report
in markdown format. Include charts if possible and save the report as
sales_analysis.md.

Web Research

# research.caw
Research the latest developments in AI coding assistants by checking
relevant websites and summarize the key findings. Focus on new features
and capabilities announced in 2024.

Meta-Programming

# extend_eagle.caw
I need a tool that can automatically optimize images - compress them,
resize them, and convert formats. It should work with JPG, PNG, and WebP files.
Create this tool and then use it to optimize all images in my /assets folder.

Eagle will:

1. Generate a complete image_optimizer tool with AI
2. Add it to your Eagle configuration automatically
3. Use the new tool to optimize your images
4. Complete the entire workflow automatically

Custom Tools

Option 1: AI-Generated Tools (Recommended)

Let Eagle write the tool for you:

# In any .caw file
Create a "weather" tool that fetches weather data for any city using a weather API.
It should take a city name and return current temperature, conditions, and forecast.

Eagle generates:

.eagle/tools/weather/
├── __init__.py     (full implementation)
├── tests.py        (unit tests)
└── README.md       (documentation)

Option 2: Manual Tool Development

For advanced users who want full control:

# .eagle/tools/my_tool/__init__.py
from eagle_lang.tools.base import EagleTool

class MyCustomTool(EagleTool):
    @property
    def name(self) -> str:
        return "my_tool"

    @property
    def description(self) -> str:
        return "Description of what my tool does"

    @property
    def usage_patterns(self) -> dict:
        return {
            "category": "custom",
            "patterns": ["Pattern 1", "Pattern 2"],
            "workflows": {"My Workflow": ["my_tool", "print"]}
        }

    def execute(self, input: str) -> str:
        return f"Processed: {input}"

💡 Note: Tools automatically update Eagle's capabilities and workflows. No restart needed!

Permission System

Tools are categorized by permission level:

• Allowed: Execute without prompts (print, read, search, call_eagle)
• Require Permission: Prompt user before execution (write, shell, web, git, ask_permission)

When a restricted tool is called, Eagle will show:

🔐 Permission Required
Tool: write
Arguments: {"file_path": "output.txt", "content": "Hello World"}
Allow this tool execution? (y/n/details):

Security & Sandboxing

Eagle includes built-in security measures:

• File Operations: Restricted to current directory and subdirectories
• Shell Commands: Blocks dangerous operations (rm -rf /, shutdown, etc.)
• Web Requests: Blocks localhost and private networks
• Git Operations: Prevents force pushes and hard resets

License

MIT License - see LICENSE file for details.

Roadmap

Building the Eagle ecosystem:

✅ Core Platform

• Interactive REPL and .caw file execution
• Self-modifying architecture with AI-generated tools
• Multi-provider AI support and project-aware configuration

🚧 Next Steps

• Better teminal experience - add colors better waiting
• Make_Eagle tool - eagle should be able to create new eagle projects with different tools
• Eagle Website - Official documentation and community hub
• Downloadable Releases - Standalone installers for all platforms
• Package Manager - eagle install <tool/config/rule> for sharing components
• VS Code Extension - Syntax highlighting and Eagle integration
• Community Registry - Discover and share tools, configs, and rules

💭 Future

• Cloud execution and team collaboration features
• Advanced debugging and workflow visualization
• Integration with popular development tools

Contributing

Eagle is in active development. Contributions, feedback, and bug reports are welcome!

The Vision

Eagle represents a new approach to AI orchestration where natural language becomes the primary interface for automation and intelligent task execution. Instead of learning syntax and frameworks, you describe what you want to accomplish.

Eagle enables:

• Faster iteration - Express ideas directly without translation to code syntax
• Lower barriers - AI orchestration accessible to anyone who can clearly explain a problem
• Direct execution - Your intent becomes reality without intermediate code generation
• True AI collaboration - Not autocomplete or suggestions, but AI agents that understand and act
• Evolving capabilities - Create custom tools and agents that grow with your needs

This is a step toward a future where the gap between human intent and computer execution continues to shrink.

---

Get started in 30 seconds:

pip install eagle-lang && eagle init && eagle

Then just start typing what you want to build.