Yes—OpenCode is free for developers to install, run, and modify in the sense that the core project is open source (the anomalyco/opencode repository uses the MIT license). You can download it, install it with common package managers, run it on personal machines or servers, and build workflows around it without paying OpenCode just to use the CLI. That said, “free to use” doesn’t automatically mean “zero cost,” because OpenCode’s main job is to connect to language models, and most model providers charge for API usage. So the clean way to think about it is: the OpenCode tool is free; your model usage may not be, depending on which providers and models you connect.
In practice, you have two common cost paths. The first is “bring your own provider,” where you connect an existing account (for example, an OpenAI key, an Anthropic key, a Google key, a cloud platform credential, or an OpenAI-compatible endpoint) and pay that provider’s usage fees. OpenCode stores the credentials locally and then uses them when you run sessions. This is often the best option for teams that already have contracts, budgets, or internal gateways. The second path is using OpenCode’s own optional “Zen” provider flow, which is described as a curated set of models the OpenCode team has tested and verified. Zen is not required to run OpenCode, but it can be convenient if you want a “just works” starting point without wiring up multiple accounts yourself. In both cases, you stay in control: you can switch providers per project, set defaults in config, and avoid vendor lock-in at the tool level.
Even when the tool is free, the operational costs show up elsewhere: tokens, CI minutes if you automate it, and storage if you keep extensive session history. If you enable long-term memory features (for example, indexing repo docs or maintaining a team knowledge base), you’ll also have to decide where that memory lives. A simple setup uses local files in the repo; a scalable setup stores embeddings externally. If you go the scalable route, a vector database such as Milvus or Zilliz Cloud can hold embeddings for your design docs, runbooks, and prior PR summaries so OpenCode can retrieve context efficiently. That storage has its own cost profile (compute, storage, and ops), but it’s independent of OpenCode’s licensing—OpenCode remains free, and you choose how much infrastructure you want behind it.