What is Amp?
Amp is an AI coding assistant for code editors and terminals. It supports complex code tasks like generation, debugging, and code review by combining multiple AI models. Users interact with Amp directly in development environments such as VS Code or the terminal. Amp uses top models like GPT-5.2 and Claude to provide accurate, task-specific responses. It also supports custom tool extensions through MCP servers.
Features & Benefits
- Multi-model orchestration – Use top-performing models like GPT-5.2, Claude Opus 4.5, and Gemini 3 across different tasks.
- Editor and terminal integration – Access AI coding assistant features inside VS Code, Neovim, JetBrains, Cursor, or directly through CLI.
- Smart and rush modes – Choose between high-performance (smart) or fast, cost-efficient (rush) AI assistance.
- Next-action code prediction – Generate in-editor suggestions using Amp-Tab-4 to continue code quickly.
- Bug detection and code review – Identify bugs with Gemini 3 Pro in the dedicated review mode.
- Codebase retrieval – Search through large projects using Gemini 3 Flash for fast and accurate results.
- Thread-based saving – Save and share AI sessions to track context, like version control for conversations.
- Media and file analysis – Analyze images, PDFs, and media files through Gemini 3 Flash.
- Custom SDK access – Build custom tools with the TypeScript SDK supporting streaming and tool configuration.
- MCP server integration – Extend Amp with custom tools by connecting local or remote MCP (Model Context Protocol) servers.
Real-world applications
A backend developer might use Amp as an AI coding assistant in VS Code to get real-time code completions with the Tab model. When fixing bugs, they could switch to the review panel powered by Gemini 3 Pro to identify logic flaws before pushing code.
Teams working across large monorepos may use the Search subagent to find references and function definitions fast, using Gemini 3 Flash. Combined with thread saving, they can keep records of key AI-guided sessions for future review.
Developers building custom workflows could set up MCP servers to link Amp with internal dev tools or specialized AI services. For instance, a local MCP server might connect to a model fine-tuned on a proprietary codebase, expanding Amp’s utility while keeping control over infrastructure.
Project leads planning high-level changes may use the Oracle mode, powered by GPT-5.2, for strategic reasoning. Amp’s librarian subagent can also help retrieve external references to back up decisions or find related open-source code.
