FETCH.AI MCP CLI

A powerful CLI for interacting with the ASI:One LLM and Model Context Protocol (MCP) servers using natural language queries - no JSON required!

Quick Start

Installation

pip install asi1-mcp-cli

📦 Available on PyPI: asi1-mcp-cli

Setup

1. Create configuration directory:

   mkdir -p ~/.asi1

2. Create configuration file:

   # Copy the example config
   cp $(python -c "import asi1_mcp_cli; import os; print(os.path.join(os.path.dirname(asi1_mcp_cli.__file__), '..', 'asi1-mcp-server-config-example.json'))") ~/.asi1/config.json

3. Edit the configuration:

   nano ~/.asi1/config.json

   Update the following fields:

   • api_key: Your ASI:One API key
   • BRAVE_API_KEY: Your Brave Search API key (optional)

Usage

Basic Conversation

asi1 "What is the capital of France?"

Web Search (Natural Language - No JSON!)

asi1 "latest developments in quantum computing"
asi1 "What are the latest developments in quantum computing?"

List Available Prompts

asi1 --list-prompts

Use Specific Prompt Templates

# Code review
asi1 run p review

# Git commit
asi1 run p commit

# YouTube video summary
asi1 run p yt https://www.youtube.com/watch?v=example

Configuration

The configuration file (~/.asi1/config.json) supports:

• api_key: Your ASI:One API key
• BRAVE_API_KEY: Your Brave Search API key (optional)
• Other LLM/MCP server settings as needed

Configuration Example

{
  "systemPrompt": "You are a helpful AI assistant",
  "llm": {
    "provider": "asi-one",
    "model": "asi1-mini",
    "api_key": "asi1-api",
    "temperature": 0,
    "base_url": "https://api.asi1.ai/v1"
  },
  "mcpServers": {
    "exa": {
      "command": "npx",
      "args": ["-y", "exa-mcp-server", "--tools=web_search"],
      "env": {
        "EXA_API_KEY": ""
      }
    },
    "brave-search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": {
        "BRAVE_API_KEY": ""
      }
    },
    "mcp-server-commands": {
      "command": "npx",
      "args": ["mcp-server-commands"],
      "requires_confirmation": ["run_command", "run_script"]
    }
  }
}

---

🔄 Workflow

graph TD
    A[User Input: Natural Language] --> B[CLI Parser]
    B --> C{Command Type?}
    C -->|Agent Chat| D[Send to Agent]
    C -->|LLM Query| E[Send to ASI:One LLM]
    C -->|MCP Query| F[Route to MCP Server]
    
    D --> G[Agent Response]
    E --> H[LLM Response]
    F --> I[MCP Server Response]
    
    G --> J[Display Result]
    H --> J
    I --> J
    
    J --> K[Auto-exit for Agent Chat]
    J --> L[Continue for LLM/MCP]
    
    style A fill:#e1f5fe
    style J fill:#c8e6c9
    style K fill:#ffcdd2

Loading

---

🚀 Features

• Natural Language Interface: Simply type your query in plain English - no JSON or complex syntax required!
• Beautiful CLI Header: Eye-catching ASCII art and welcome message every time you start the CLI.
• Agent Chat: Send messages to other agents using asi1 agent chat --address <agent_address> --message "...".
• LLM/MCP Queries: Use asi1 <query> or asi1 run <query> for natural language queries to the LLM or MCP server.
• Prompt Templates: Use pre-defined prompt templates for code review, commit messages, and YouTube summaries.
• Backwards Compatible: Old usage patterns still work—no need to change your workflow!
• Call Agent, MCP Server, and LLM: The CLI can route your query to an agent, the MCP server, or the LLM as needed.

---

🖥️ Example Usage

1. Show the CLI header and help

asi1 --help

2. Ask the LLM or MCP server a question (Natural Language!)

asi1 "What is the capital of France?"
asi1 "latest developments in quantum computing"
asi1 "Build a chat bot in JavaScript and HTML"
asi1 run "What are the latest AI developments?"

3. Use a prompt template

asi1 run p review
asi1 run p commit
asi1 run p yt https://www.youtube.com/watch?v=dQw4w9WgXcQ

4. Chat with another agent

asi1 agent chat --address <agent_address> --message "Get the latest tweet from Elon Musk"

---

🧠 Prompt Templates

You can use prompt templates for common tasks:

• review: Code review for your current git diff
• commit: Generate a Conventional Commit message for your changes
• yt: Summarize a YouTube video by URL

Example:

asi1 run p yt https://www.youtube.com/watch?v=dQw4w9WgXcQ

---

🔗 External Resources

• PyPI Package: Official package on Python Package Index
• ASI:One Documentation: Official ASI:One platform documentation
• Agentverse: Discover and interact with AI agents
• ASI:One Models: Choose from asi1-mini, asi1-fast, asi1-extended, asi1-agentic, and asi1-graph

---

🆕 Recent Improvements

• Natural Language Interface: No more JSON! Just type your query in plain English
• Premium, colorful CLI header with FETCH.AI MCP CLI branding
• Agent chat with auto-exit after first response
• Backward compatibility for LLM/MCP queries
• Improved help and usage examples
• Prompt templates for review, commit, and YouTube summary

---

🤝 Get Help

• Run asi1 --help for usage and examples
• For issues, open an issue on GitHub or contact the maintainers

---

No more complex JSON! Just type your query naturally and let the CLI handle the rest.

Enjoy using the FETCH.AI MCP CLI! Powered by fetch.ai and ASI:One.