OpenCode is a terminal-native coding agent built by Anomaly Innovations — the team behind the open-source AWS framework SST — and released under the MIT license. Unlike Claude Code or GitHub Copilot, it isn't tied to one model vendor. It connects to more than 75 AI providers through the AI SDK and Models.dev, covering Anthropic, OpenAI, Gemini, DeepSeek, OpenRouter, and fully local setups through Ollama or LM Studio. It ships two operating modes worth knowing about: Build mode, which gives the agent full read/write/execute access to your codebase, and Plan mode, a read-only pass meant for understanding a repo before you let it touch anything.

Installation is a single command (`curl -fsSL https://opencode.ai/install | bash` on the official docs, or via your package manager of choice), and because the whole project is MIT-licensed, self-hosting is not a workaround — it's the intended default. You can inspect the source, fork it, or run it fully air-gapped if your org needs that. That's a meaningfully different trust model than a closed agent where you're taking the vendor's word for what happens to your code.

Here's the part that actually explains the growth curve: in June 2026, SpaceX confirmed a roughly $60 billion acquisition of Cursor's parent company — reportedly the largest startup acquisition on record — and Cursor's market share reportedly slid from 41% to 26% over the following year. Developers who liked Cursor's editor experience but didn't want their code flowing through a rocket company's infrastructure needed somewhere to go. OpenCode's pitch — model-agnostic, vendor-neutral, terminal-first — turned out to be exactly the alternative that moment called for. Momentum like that doesn't always hold, but the timing here isn't a coincidence, and it's the honest explanation for why an open-source CLI tool is currently out-starring every closed competitor.

The "free" label needs an asterisk, though, and it's worth being direct about it. OpenCode itself costs nothing — but you're paying for whichever model you point it at. Route it through Claude Opus 5 or GPT-5.6 Sol at frontier pricing and your monthly bill can land close to, or even above, what Claude Code's flat $20–$200/month subscription would have cost you, especially if your usage is heavy and unpredictable. Where OpenCode actually wins on cost is when you either run local models through Ollama at zero marginal cost, or deliberately mix cheaper models for routine tasks and save the expensive ones for hard problems — something a single-vendor subscription tool can't do for you.

So is it worth switching without the safety net of a flat monthly price? If model lock-in genuinely bothers you, or your team needs to self-host for compliance reasons, yes — there's no closed competitor offering the same flexibility. If you just want predictable pricing and don't care which model is under the hood, a managed subscription tool still removes more friction than OpenCode will, at least until you've tuned your provider mix.