Developers looking for AI-powered coding assistance without vendor lock-in now have a promising new option. Sidekick, currently in beta, offers an agentic CLI-based alternative to proprietary tools like Claude Code, GitHub Copilot, and Cursor, with the key advantage of supporting multiple large language model providers.
Flexibility Across LLM Providers
Sidekick stands out from its competitors by allowing developers to choose their preferred LLM provider rather than being tied to a single AI model. The tool currently supports all major models including OpenAI (with GPT-4.1), Anthropic's Claude, and Google's Gemini models. This flexibility addresses a common pain point for developers who want the benefits of AI assistance without committing to a single ecosystem.
I think Aider is great, but when I first tried it, it wasn't agentic - you had to manually manage files in the context window. I found this cumbersome. I definitely resonated with Claude Code more, but wanted more model options.
Agentic Workflow with Safety Features
The tool's agentic workflow means it can autonomously perform coding tasks with minimal manual intervention. One particularly appreciated feature is the /undo command, which allows developers to quickly revert changes when the AI makes mistakes. This safety net encourages experimentation while reducing the risk of breaking code. Sidekick also includes a yolo mode that skips tool confirmations for developers who prefer a more streamlined experience.
Key Features of Sidekick
- Multiple LLM provider support (OpenAI, Anthropic, Google)
- MCP (Model Context Protocol) support
- "/undo" command for reverting AI changes
- Easy model switching within sessions
- Per-project guide customization
- CLI-first design
- Cost and token tracking
- Configurable confirmation settings
Installation Methods
- Using pip:
pip install sidekick-cli - From source: Clone repository and install dependencies
Available Commands
- /help - Show available commands
- /yolo - Toggle "yolo" mode (skip confirmations)
- /clear - Clear message history
- /compact - Summarize history and clear old messages
- /model - List available models
- /model - Switch to specific model
- /undo - Revert recent changes
- /dump - Show message history
- /exit - Exit application
Customization and Context Management
Sidekick supports project-specific customization through a SIDEKICK.md guide file that provides instructions about tech stack, project structure, and development preferences. This helps maintain context across coding sessions. The tool also supports Model Context Protocol (MCP) servers, extending its capabilities to interact with additional tools and data sources.
For developers concerned about context management, Sidekick offers commands like /compact to summarize message history and clear old messages, helping maintain focus during extended coding sessions.
While still in beta with some quirks to iron out, Sidekick shows promise as an open-source alternative in the growing field of AI coding assistants. The creator mentions that local model support is on the roadmap but has focused initially on cloud providers that have shown the most promise for coding tasks. As development continues, the community will be watching to see if Sidekick can deliver the same quality of assistance as its proprietary counterparts while maintaining its open and flexible approach.
Reference: Sidekick (Beta)