Crush: Your Terminal's AI Coding Companion

Elevate Your Terminal Workflow with Crush: The Ultimate AI Coding Companion

In the fast-paced world of software development, efficiency and seamless integration are paramount. Meet Crush, a groundbreaking open-source project that brings the power of Large Language Models (LLMs) directly into your favorite terminal environment. Billed as "the glamourous AI coding agent for your favourite terminal 💘," Crush is designed to be your new coding best friend, wiring your tools, code, and workflows directly into your chosen LLM.

What is Crush?

Crush is a sophisticated AI agent that resides within your terminal, offering a powerful interface to interact with various AI models. It's built with the developer in mind, providing a flexible, session-based, and extensible platform to enhance your coding productivity. Whether you're writing code, debugging, or exploring new ideas, Crush aims to streamline the process by bringing AI assistance right where you work.

Key Features That Set Crush Apart:

• Multi-Model Support: Crush doesn't tie you to a single AI provider. It allows you to choose from a wide array of LLMs or even integrate your own custom OpenAI- or Anthropic-compatible APIs. This flexibility ensures you can leverage the best model for the task at hand.
• Flexible Switching: Need to switch LLMs mid-session? Crush allows you to do so while preserving your context, ensuring a fluid and uninterrupted workflow.
• Session-Based Context: Manage multiple work sessions and contexts per project. This feature is invaluable for keeping track of different tasks and maintaining focus.
• LSP-Enhanced: Crush intelligently uses Language Server Protocol (LSP) for richer context, mirroring how developers themselves often work with code.
• Extensibility: With support for Model Context Protocol (MCP) servers via HTTP, stdio, and SSE, Crush is highly extensible, allowing you to integrate custom capabilities.
• Cross-Platform Compatibility: Crush works everywhere. It boasts first-class support on macOS, Linux, Windows (including PowerShell and WSL), FreeBSD, OpenBSD, and NetBSD, making it accessible to a vast range of users.

Getting Started with Crush

Installation is straightforward. Crush is available through various popular package managers like Homebrew, NPM, and Nix. You can also install it directly using Go or by downloading pre-compiled binaries. For detailed instructions tailored to your operating system, consult the official documentation.

To begin using Crush, you'll typically need an API key for your preferred AI provider, such as Anthropic, OpenAI, Groq, or OpenRouter. You can either enter this key when prompted or set it via environment variables like OPENAI_API_KEY, ANTHROPIC_API_KEY, or GROQ_API_KEY.

Customization and Configuration

Crush is designed to work out-of-the-box, but it also offers extensive customization options. You can configure Crush locally within your project or globally. Configuration is managed through JSON files (e.g., .crush.json, crush.json, $HOME/.config/crush/crush.json), allowing you to define LSPs, MCP servers, whitelist tools, and even set up custom providers.

The project also includes helpful tools for managing logs, with commands like crush logs to view recent activity or crush logs --follow to monitor them in real-time. For deeper insights, you can enable debug modes through configuration or by using the --debug flag.

Contributing to the Ecosystem

Crush's open-source nature encourages community involvement. The project welcomes contributions, whether it's adding support for new AI models through the Catwalk repository or improving the core functionality. If you have ideas or need assistance, the Charm community is active on Discord, Twitter, and other platforms.

Crush represents a significant step forward in integrating advanced AI capabilities into the developer's daily workflow. By bringing powerful LLMs directly into the terminal, it offers a more focused, efficient, and personalized coding experience. If you're looking to supercharge your productivity and explore the future of AI-assisted development, Crush is an open-source project well worth your attention.