Welcome to CodeCompanion.nvim

AI Coding, Vim Style

CodeCompanion lets you code with LLMs and agents in Neovim.

Features

• 💬 Copilot Chat meets Zed AI, in Neovim
• ⚡ Run CLI agents like Claude Code and Codex alongside your code
• 🔌 LLMs from Anthropic, DeepSeek, Gemini, GitHub Copilot, Hugging Face, Kimi, Mistral, Novita, Ollama, OpenAI, Azure OpenAI, OpenRouter and xAI out of the box, or bring your own
• 🤖 Agents over the Agent Client Protocol, including Auggie, Cagent, Claude Code, Cline, Codex, Copilot CLI, Cursor CLI, Gemini CLI, Goose, Kilo Code, Kimi CLI, Kiro, Mistral Vibe and OpenCode
• 🫶 Community adapters, contributed and supported by users
• 👨‍💻 Code reviews to comment on, accept or revert an agent's changes
• 🔋 Model Context Protocol (MCP) servers
• 🚀 Inline code creation and refactoring
• 🤖 Editor context, slash commands, tools and workflows to improve the LLM's output
• 🧠 Skills and rules, such as CLAUDE.md, .cursor/rules and your own
• ✨ A built-in prompt library for common tasks, like explaining code or LSP errors
• 🏗️ Your own prompts, editor context and slash commands
• 📥 Multiple chats open at the same time
• 🎨 Images and PDFs as input
• 💪 Async execution, so Neovim stays responsive

Overview

CodeCompanion is built around interactions, the different ways you work with an LLM. The chat interaction is a buffer where you converse with an LLM or an agent, and the inline interaction writes the LLM's response straight into the buffer you're editing.

An adapter connects an interaction to an LLM or an agent, and sets its model and parameters. Each interaction, and each prompt library item, can use a different adapter. The full set is in the adapters folder. To write your own, see Extending with Adapters.

Using the Docs

Ask an LLM to search these docs for you with the @{search_help} tool. On the site, use the search bar at the top left of the page.