OpenCode Superapp

Product Introduction

1. Definition: The OpenCode Superapp is a native, local-first AI workspace for macOS, built on the open-source OpenCode runtime. It is a desktop application that functions as an integrated development environment (IDE) and agentic AI platform, enabling users to interact with AI models through a unified interface that includes code editing, project management, file systems, Git, terminals, and specialized AI-powered tools.
2. Core Value Proposition: It exists to provide developers and technical users with a private, extensible, and model-agnostic workspace that brings advanced "Computer Use" and agentic capabilities—inspired by systems like OpenAI's Codex—to locally run or self-hosted AI models. Its primary value is granting users powerful AI assistance without locking them into a specific cloud provider, subscription model, or compromising data privacy.

Main Features

1. Multi-Provider, Model-Agnostic AI Runtime: The Superapp acts as a front-end for the OpenCode runtime, allowing users to connect to and switch between various AI model providers on a per-task basis. This includes major cloud APIs (OpenAI GPT-5.6, Anthropic Claude, Google Gemini, xAI), locally managed models via an integrated Ollama engine, and self-hosted OpenAI-compatible endpoints. User credentials and billing remain with their chosen providers.
2. Supervised Native Computer Use: A core "Superpower" feature that allows the AI agent (Supe) to safely inspect and operate native macOS applications. This is enabled through guarded tools and requires macOS 15 or newer. It moves beyond chat-based interaction to allow direct, agent-driven manipulation of desktop software under user supervision.
3. Integrated Local AI & Private Dictation: The app includes an app-managed Ollama engine for running local large language models (LLMs) in a private loopback network, discovering models from a user's existing library. Coupled with this is offline, multilingual speech recognition for private dictation, allowing prompt composition and control via voice without cloud processing.
4. Extensible Workspace with Skills & MCPs: The workspace is designed for extensibility through a public extension contract. Users can install or create Commands, right-panel apps, Skills, Model Context Protocol (MCP) services, and local resource packages. A unique feature is the ability to describe and create custom Home widgets with live previews, which are project-aware and have bounded file permissions.
5. Secure Remote & Realtime Voice Session: Users can pair a mobile device or browser to their Mac host to continue a live session remotely. The connection is end-to-end encrypted via a relay that cannot decrypt content. This works alongside "Supe Realtime Voice," enabling conversational, real-time voice interaction to create and manage tasks, using either a limited cloud trial or a user's configured voice provider.

Problems Solved

1. Pain Point: Vendor lock-in and privacy concerns with cloud-based AI coding assistants. Many AI development tools are tied to a single provider's models and cloud infrastructure, raising costs, limiting choice, and sending sensitive code and project data to third-party servers.
2. Target Audience: Privacy-conscious software engineers, developers working with proprietary or regulated codebases, AI researchers and hobbyists experimenting with local LLMs, and power users who want deep integration of AI into their desktop workflow beyond a simple chat window.
3. Use Cases: A developer can use a local model like Qwen 3 for routine code completion and refactoring on a private project, then seamlessly switch to a cloud model like GPT-5.6 for complex problem-solving, all within the same session. They can then use the "Computer Use" Superpower to have the agent automate a repetitive task in another Mac app, or use "Remote" to continue debugging from their phone without interrupting the agent's state.

Unique Advantages

1. Differentiation: Unlike monolithic cloud IDE assistants (e.g., GitHub Copilot embedded in VS Code) or standalone AI chat applications, OpenCode Superapp decouples the AI runtime from the interface and the model provider. It offers a more integrated, native desktop experience than web-based tools and provides advanced agentic capabilities (Computer Use, Browser Automation) typically not available in open-source or local-first alternatives.
2. Key Innovation: Its "local-first by default" architecture combined with a commercial one-time purchase model for advanced capabilities. The app requires no account, stores all runtime data and projects locally, and offers a lifetime license for its "Superpowers" (Browser Automation, Computer Use, Dictation, Remote, Realtime Voice). This contrasts sharply with the prevalent SaaS subscription model for similar advanced AI features.

Frequently Asked Questions (FAQ)

1. Is OpenCode Superapp from OpenAI? No, OpenCode Superapp is an independent, Codex-inspired product built on the open-source OpenCode runtime. It is not affiliated with, endorsed by, or a product of OpenAI.
2. Can I use OpenCode Superapp completely offline? Yes, the core application and several Superpowers like Browser Automation, Native Computer Use, and Local Dictation are designed for indefinite offline use. Local AI models run fully on your Mac. The Remote feature requires an internet connection for the encrypted relay, and Realtime Voice depends on its configured provider (which can be local).
3. What are the system requirements for OpenCode Superapp? The initial release requires an Apple Silicon Mac (M-series chip) running macOS 13 (Ventura) or newer. The Native Computer Use Superpower specifically requires macOS 15 (Sequoia) or newer due to its underlying accessibility APIs.
4. How does the pricing model work? The OpenCode Superapp base application is free forever and includes multi-provider model support, local AI, the extensible workspace, and widgets. A one-time purchase of $49 unlocks lifetime access to the "Superpowers" (Computer Use, Browser Automation, etc.) for up to five Mac installations, with no subscription.
5. How does data privacy and security work? The app is private by design: no accounts, no behavioral analytics, no cookie tracking. Projects and local model traffic stay on your Mac. Cloud model requests go directly to your chosen provider. The Secure Remote feature uses end-to-end encryption, meaning the session relay cannot decrypt your data.