SHELL-AI

What is SHELL-AI?

Shell AI is a powerful, intelligent command-line shell assistant that uses local AI models (powered by Ollama) directly in your terminal. It enables natural language interaction with your file system, shell commands, and development workflow, with all processing happening offline for privacy.

How to use SHELL-AI?

Install globally with npm install -g @shell-ai/cli or run immediately with npx @shell-ai/cli. You need Ollama installed and a model pulled (e.g., ollama pull gpt-oss). Start an interactive chat with shell-ai or run shell-ai --model mistral to use a specific model. Configure MCP servers by creating ~/.shell-ai/mcp.json.

Key features of SHELL-AI

• Intelligent ReAct agent reasons and uses tools for complex tasks
• Model Context Protocol (MCP) integration for custom services
• Zero-network inference via Ollama – fully offline and private
• Extensive built-in toolset: file ops, shell exec, web search, memory
• Interactive log viewer with real-time filtering and navigation
• Fully typed TypeScript/TS 5.x support for robust development
• Modular workspace architecture separating core and CLI packages

Use cases of SHELL-AI

• Manage files and directories using natural language commands
• Execute shell commands safely (with user confirmation) through AI
• Perform web research and Wikipedia lookups without leaving the terminal
• Store and retrieve long-term memory facts for repeated tasks
• Connect to external APIs and databases via MCP servers for custom workflows

FAQ from SHELL-AI

How is Shell AI different from cloud-based AI assistants?

Shell AI runs entirely on your machine using Ollama, with no network calls for inference. It uses a ReAct agent that can reason and call local tools, keeping your data private and enabling offline operation.

Which AI models does Shell AI support?

It supports any model you have pulled through Ollama, such as mistral, llama3, or gpt-oss. Specify the model with the --model flag when starting the session.

How do I extend Shell AI with MCP servers?

Create a configuration file at ~/.shell-ai/mcp.json listing your MCP servers (HTTP, SSE, or STDIO transport), including authentication methods like Bearer token, Basic Auth, or OAuth2. Tools from connected servers are automatically added to the agent’s tool registry.

Is Shell AI free to use?

Yes, Shell AI is open source under the Apache‑2.0 license. You only need Ollama (free) and a pulled language model.

Does Shell AI require an internet connection?

Core inference and file operations work offline. However, the web-search and wikipedia-search tools do require internet access when used.