OpenCode
OpenCode is an MIT-licensed coding agent for the terminal, IDE, and desktop. The GitHub repo is now anomalyco/opencode (formerly under sst). On 2026-08-17 the API listed about 198,202 stars; the homepage claims 195k stars, 950 contributors, 13,000+ commits, and 16M monthly developers. Official copy says it does not store your code or context. You can use included free models, connect Claude / GPT / Gemini / 75+ providers via Models.dev, or log in with GitHub Copilot or ChatGPT Plus/Pro.
Key Features
- Three surfaces: Terminal UI, desktop beta (macOS, Windows, Linux, with tabs), and IDE extension.
- Any model: Homepage lists 75+ providers including local models, plus Copilot and ChatGPT logins.
- LSP, multi-session, share links: Loads language servers, runs parallel agents on one project, and can share a session link.
- Zen: A curated, benchmarked model set for coding agents so you do not shop raw providers.
- Privacy-first claim: "OpenCode does not store any of your code or context data." Confirm that against the current privacy page before a regulated rollout.
Use Cases
- Terminal-native teams that want an open agent instead of Claude Code or Codex CLI.
- Bring-your-own-key shops that already pay Copilot or ChatGPT and want one TUI.
- Desktop beta when you want tabs on Mac/Windows/Linux without leaving the OpenCode project.
Pricing
The agent itself is open source (MIT). Homepage FAQ asks "How much does OpenCode cost?" and advertises free included models. Provider usage (Claude, GPT, Gemini, Copilot, ChatGPT) is billed by those vendors. Zen is a paid/curated model access layer; read opencode.ai for the current Zen offer. Do not invent a seat price.
Getting Started
- Install with
curl -fsSL https://opencode.ai/install | bash(npm, bun, brew, paru are also listed). - Open a repo and run OpenCode. Add provider keys or sign in with Copilot / ChatGPT.
- Read the docs intro for configure, LSP, MCP, and permissions.
- Try one session share link internally before you treat it as a debug default.
Frequently Asked Questions
Is it still the sst repo?
The live GitHub canonical is anomalyco/opencode. sst redirects there.
Do I need another AI subscription?
Not strictly. Free included models exist. Copilot and ChatGPT logins are optional.
Terminal only?
No. Docs and homepage list desktop and IDE as well.
Alternatives
- Claude Code: Anthropic's official multi-surface agent.
- Codex CLI: OpenAI's ChatGPT-billed terminal agent.
- Continue: Open IDE agent after the Cursor acquisition.
Tips
- Pin the GitHub org (
anomalyco) in internal docs so install scripts do not hit a stale sst path. - Keep secrets in provider config, not in shared session links.
- Recheck star and MAU claims. The API star count and the homepage marketing number already differ by a few thousand.
Conclusion
OpenCode is the high-star MIT coding agent to try when you want terminal, IDE, and desktop without locking to one lab. Install from opencode.ai, then decide whether free models, Zen, or an existing Copilot/ChatGPT login is enough.
Comments
No comments yet. Be the first to comment!
Related Tools
Junie (JetBrains)
junie.jetbrains.com
JetBrains' LLM-agnostic AI coding agent that ships code from your terminal, IDE, or CI/CD pipeline, with bring-your-own-key model support.
Continue
continue.dev
Open-source coding agent for VS Code and JetBrains. Apache-2.0, local models, CLI, and MCP. Acquired by Cursor; the repo stays public.
Kilo Code
kilo.ai
Open-source VS Code/CLI agent. Rechecked: individuals Free $0; Teams $15; Enterprise custom. BYOK or Kilo Gateway. Not a 1.4M-install census.
Related Insights
Hook OpenCode Go into Codex on Windows. Do Not Open a Second Toolkit.
A ChatGPT-signed Codex desktop app still shows mostly GPT in the picker. On Windows, enable only OpenCode Go and the Grok, GLM, Kimi, DeepSeek, and MiniMax models you already pay for appear in the same selector. Keys stay local. Native GPT stays put.
What Locks Codex Is the Picker, Not the Models
You already pay for OpenCode Go, Grok, and Z.ai, but the Codex picker still shows mostly GPT. The community project codex-router does not teach another install ritual. It puts subscriptions you already bought back into the selector, keeps keys on the machine, and leaves native GPT alone.
Seven AI Coding CLIs, Six Months: No Matter How Strong the Model, Work Needs Supervision
Claude Code, Codex, opencode, pi, omp and DeepSeek Harness all have personalities. After six months of deep use I run a division of labor: pi for the fastest cheapest reviews, omp for complex PRs, DeepSeek Harness on V4 Flash for high-frequency low-cost review, and Claude Code, Qoder and Cursor for writing. No matter how strong the model, work needs supervision — ideally from an independent third party.