OpenCode is an open source AI coding agent you can run in your terminal, as a desktop app, or as an IDE extension. You describe a task, and it plans and edits code with the model you pick. It connects to 75+ LLM providers through Models.dev, including local models. You can also sign in with an existing GitHub Copilot or ChatGPT Plus/Pro account. If you want to choose your own models, keep your code private, and work mostly in the terminal, OpenCode is free and you pay only for model usage.

• 75+ LLM Providers: You can connect Claude, GPT, Gemini, local models, and many more through Models.dev. You aren't tied to one vendor.
• Use Your Existing Subscriptions: Sign in with GitHub Copilot or ChatGPT Plus/Pro and use plans you already pay for.
• Automatic LSP Integration: OpenCode loads the right language servers for your project automatically, so the model gets real code intelligence instead of guessing.
• Plan and Build Modes: Press Tab to switch modes. Plan mode suggests changes without touching your files, and Build mode makes the edits.
• Multi-Session Agents: You can run several agents in parallel on the same project and split work across separate tasks.
• Shareable Sessions: The /share command creates a link to a conversation that you can use for reference or debugging. Sessions are private unless you share them.
• Terminal, Desktop, and IDE: You get the same agent as a terminal UI, a desktop app, or an IDE extension, plus GitHub and GitLab integrations.
• Extensible Setup: You can add MCP servers, custom tools, agents, agent skills, plugins, themes, keybinds, and formatters. A project AGENTS.md file holds your project rules.

• Free and open source: The agent itself costs nothing, and the code is public on GitHub. The project has over 200,000 stars and about 950 contributors.
• Model freedom: You can switch providers or run local models whenever you like. Reviewers often name this as its main advantage over single-vendor agents.
• Fits existing workflows: It runs in your terminal next to your current editor, so you don't have to switch IDEs.
• Flexible paid options: Use your own API keys, access supported free models, or choose OpenCode's paid Go subscription for access to selected coding models.
• Large, active community: OpenCode reports more than 16 million developers using the tool each month, with an active open-source project and community.

• Model costs are separate: The agent is free, but you pay for whatever models you use through API keys, Zen credits, or a Go plan. Your total cost depends on how much you use it.
• Terminal-first experience: OpenCode centers on a terminal workflow. A desktop app and IDE extension exist, but if you want an AI-native editor with visual diff review built in, a full IDE like Cursor may suit you better.
• Native Windows support is still in progress: The docs recommend running OpenCode through WSL on Windows, which adds a setup step for Windows-only teams.
• Enterprise pricing isn't public: The Enterprise plan adds SSO and internal AI gateway support, but you have to contact sales for a quote, which slows budget planning.

• Building features from the terminal: Describe a new endpoint or component, review the proposal in Plan mode, then switch to Build mode to apply it. You stay in your terminal and editor the whole time.
• Refactoring large codebases: LSP integration gives the agent real code intelligence for symbols and references. Running parallel sessions lets you split a big refactor into separate tasks.
• Team debugging and knowledge sharing: Share a session link with a teammate so they can see exactly how the agent reached a fix. It speeds up code review and onboarding.
• Privacy-sensitive and enterprise development: Use local models or an internal AI gateway with enterprise SSO to keep code and context within your chosen environment. Review sharing settings and your AI provider's data policies.