All posts
announcement open-source

Introducing ChatShell: An AI Agent Ready Out of the Box

ChatShell Team ·

We built ChatShell because we were tired of the same pattern: install an AI chat app, realize it can’t actually do anything without hours of configuration, then spend more time configuring plugins than using the app.

ChatShell is different. The moment you install it, you have an AI agent that can search the web, fetch web pages, execute shell commands, read and edit files on your machine — with no plugins to install and no MCP servers to configure.

What Makes ChatShell Different

Most AI desktop clients are glorified API wrappers. They give you a chat interface, let you pick a model, and call it a day. If you want the AI to actually take actions — search the web, run code, browse a page — you’re on your own.

ChatShell ships with 9 built-in tools that your AI uses autonomously:

  • Web Search — Searches the web and retrieves relevant results
  • Web Fetch — Fetches the full content of any web page
  • Bash — Executes shell commands on your machine
  • Read — Reads files from your local filesystem
  • Edit — Makes precise text replacements in files
  • Write — Creates or overwrites files with provided content
  • Grep — Searches file contents with pattern matching
  • Glob — Discovers files by name patterns
  • Kill Shell — Terminates the current bash session and starts fresh

These aren’t afterthoughts. They’re designed to work together seamlessly. Here’s what that looks like in practice:

  • Travel Planning with Web Search — AI searches the web, fetches travel guides, and creates a detailed 3-day itinerary, all autonomously.
  • Movie Research with Deep Web Fetching — 18 tool calls in one query: 6 web searches + 12 page fetches to find and compare the best sci-fi movies.
  • Project Analysis with Local Tools — 20 tool calls using Glob, Bash, and Read to analyze a codebase and produce a structured architecture overview.

Beyond the Built-in Tools

ChatShell ships with more than just tools. It’s a complete AI agent platform:

Skills bundle prompt instructions with required tools into reusable, composable capabilities. Create your own by placing a SKILL.md file inside its own subdirectory under ~/.chatshell/skills/<skill-name>/SKILL.md. Each skill declares which tools it needs — those tools are enabled automatically when the skill is active.

Custom Assistants let you create personalized AI personas entirely on your machine. Each assistant packages a model, system prompt, tools, skills, and a custom avatar into one reusable identity you can switch between instantly.

MCP Integration connects ChatShell to external tool servers via STDIO or HTTP. Supports OAuth 2.0/2.1 with PKCE for enterprise auth, per-conversation server selection, and dynamic tool discovery.

40+ AI Providers — OpenAI, Anthropic, Gemini, DeepSeek, Ollama, and many more cloud and local providers. All with automatic model discovery.

Built on Tauri 2 + Rust

We chose Tauri 2 with a Rust backend deliberately. Electron apps are heavy. ChatShell has a low memory footprint, fast startup time, and native performance — on macOS, Windows, and Linux.

Your conversations are stored locally in SQLite. API keys and MCP tokens are encrypted with AES-256-GCM, with the master key stored in the OS keychain. There’s no telemetry and no cloud dependency.

Open Source, Apache 2.0

ChatShell is open source under the Apache 2.0 license — no GPL restrictions, no per-seat pricing. Use it however you want.

The source code is available on GitHub.

Try It Today

ChatShell is available for macOS, Windows, and Linux. Head to our download page to get the latest version, or browse the source code to see how it works.

Questions or feedback? Open an issue on GitHub — we’d love to hear from you.