You want an open-source coding agent that does not lock you into one model vendor. OpenCode and Goose both qualify: free, BYO-key, terminal-native. They differ in scale and scope. OpenCode has 172,198 GitHub stars (MIT) and 75+ model providers, the most-starred open-source coding agent. Goose has 48,542 stars (Apache-2.0), is written in Rust, now lives under the Linux Foundation, ships a desktop app, and handles more than code.
Summary
| Dimension | OpenCode | Goose |
|---|---|---|
| GitHub stars | 172,198 | 48,542 |
| License | MIT | Apache-2.0 |
| Repo | anomalyco/opencode | aaif-goose/goose |
| Governance | Community (anomalyco) | Linux Foundation AAIF |
| Surfaces | Terminal TUI | Desktop app + CLI + API |
| Model providers | 75+ via AI SDK | 15+ providers |
| Built in | TypeScript / AI SDK | Rust |
| Scope | Coding agent | Code + research + writing + automation |
| Price | Free, BYO key | Free, BYO key |
Stars, License, and Governance
OpenCode is the most-starred open-source coding agent on GitHub at 172,198 stars, ahead of gemini-cli (105,104) and openai/codex (89,991). The repo moved to anomalyco/opencode (the old sst/opencode URL redirects), and it ships under the MIT license. OpenCode Zen is the team's curated list of models tested and verified for agentic coding.
Goose has 48,542 stars under Apache-2.0. The notable 2026 change is governance: Goose moved from block/goose to the Agentic AI Foundation (AAIF) at the Linux Foundation, now at aaif-goose/goose, with a public GOVERNANCE.md and support for custom distributions. If neutral, foundation-backed governance matters for your stack, Goose has it; OpenCode is driven by anomalyco.
Most-starred open-source coding agent
Install Commands
| Method | OpenCode | Goose |
|---|---|---|
| Shell script | curl -fsSL https://opencode.ai/install | bash | curl -fsSL .../aaif-goose/goose/releases/download/stable/download_cli.sh | bash |
| npm | npm install -g opencode-ai | Not published |
| Homebrew | brew install anomalyco/tap/opencode | Desktop app download |
| Other | Bun, pnpm, Yarn, pacman, Scoop, Docker | Desktop app for macOS / Linux / Windows |
OpenCode ships through more package managers (npm, Bun, pnpm, Yarn, pacman/paru, Chocolatey/Scoop/Mise, Docker). Goose distributes the CLI via a release script and a native desktop app for macOS, Linux, and Windows.
Model Providers
OpenCode supports 75+ LLM providers through the AI SDK and the Models.dev catalog, plus local models via Ollama, LM Studio, and llama.cpp. Custom OpenAI-compatible providers are configured in JSON:
{"provider":{"myprovider":{
"npm":"@ai-sdk/openai-compatible",
"options":{"baseURL":"https://api.myprovider.com/v1"},
"models":{ ... }
}}}Goose works with 15+ providers (Anthropic, OpenAI, Google, Ollama, OpenRouter, Azure, Bedrock, and more) and connects to 70+ extensions over MCP. OpenCode has the wider model menu; Goose has the broader extension ecosystem and a general-purpose scope: research, writing, automation, and data analysis, not code alone.
Subscription Reuse
A common question is whether you can point either tool at a Claude Pro or Max plan instead of paying per token. You cannot. OpenCode's docs state Anthropic explicitly prohibits using Claude Pro/Max subscriptions with third-party tools like OpenCode. Both tools accept a paid Anthropic API key.
| Subscription | OpenCode | Goose |
|---|---|---|
| Claude Pro / Max | Prohibited by Anthropic | Reusable via ACP |
| ChatGPT Plus | Usable as backend | Reusable via ACP |
| GitHub Copilot | Usable as backend | Not documented |
| GitLab Duo | Usable as backend | Not documented |
| Gemini subscription | Not documented | Reusable via ACP |
Goose reuses existing Claude, ChatGPT, and Gemini subscriptions through ACP. OpenCode reuses ChatGPT Plus, GitHub Copilot, and GitLab Duo subscriptions as model backends. If you live on a Claude subscription and want a third-party agent, Goose's ACP path is the documented route; OpenCode points you to a paid Anthropic API key.
Where OpenCode Wins
Largest community
172,198 stars, the most-starred open-source coding agent. Daily pushes.
75+ providers
AI SDK plus Models.dev, OpenAI-compatible JSON config, local via Ollama / LM Studio / llama.cpp.
MIT license
Permissive license and broad package-manager coverage for embedding in your stack.
Where Goose Wins
Desktop app
GUI on macOS, Linux, and Windows alongside the CLI and API. A non-terminal option.
Foundation governance
Linux Foundation AAIF with a public GOVERNANCE.md and custom-distribution support.
Beyond code
Built for research, writing, automation, and data analysis, with 70+ MCP extensions.
Which Model to Run Inside Each
Both agents are BYO-model, so the model you connect sets your quality and cost. The frontier coding models lead the benchmarks: Claude Opus 4.8 scores 88.6% on SWE-bench Verified (self-reported, llm-stats), Codex CLI with GPT-5.5 tops Terminal-Bench 2.1 at 83.4%, and Claude Code with Opus 4.8 follows at 78.9%. Those run at frontier API prices: Opus 4.8 is $5/M input and $25/M output, GPT-5.5 is $5/M input and $30/M output.
For open-source models, where you serve them changes both quality and cost. Most serverless providers quantize activations to fp8 to cut cost, which degrades output. Morph serves DeepSeek with 16-bit (bf16) activations and no fp8/int8 quantization, so responses match the reference weights. That makes it the best place to run DeepSeek when output fidelity matters. For coding specifically, Morph runs codegen-tuned speculative decoding (draft/ngram tuned on code) plus custom low-level inference kernels, which makes it the fastest and highest-quality option for coding agents rather than a general-purpose menu.
| Model | Price (input / output per 1M) | Note |
|---|---|---|
| Claude Opus 4.8 | $5 / $25 | 88.6% SWE-bench Verified (self-reported) |
| GPT-5.5 | $5 / $30 | #1 Terminal-Bench 2.1 (83.4%) |
| morph-dsv4flash (DeepSeek V4 Flash) | $0.139 / $0.278 | 16-bit activations, codegen-tuned serving |
morph-dsv4flash runs at $0.139 per 1M input tokens and $0.278 per 1M output tokens, served at full 16-bit precision. See Morph Open Source Models and pricing.
Decision Framework
| Your priority | Best choice | Why |
|---|---|---|
| Largest community / most stars | OpenCode | 172,198 stars, daily development. |
| Widest model menu | OpenCode | 75+ providers via AI SDK and Models.dev. |
| Desktop GUI | Goose | Native app on macOS, Linux, Windows. |
| Foundation governance | Goose | Linux Foundation AAIF with public GOVERNANCE.md. |
| Non-code tasks (research, writing) | Goose | General-purpose by design. |
| Reuse a Claude subscription | Goose | Reusable via ACP (OpenCode requires API key). |
| Reuse Copilot or GitLab Duo | OpenCode | Documented as usable backends. |
Frequently Asked Questions
Is OpenCode or Goose better?
OpenCode for a terminal-first coding agent with the largest community (172,198 stars) and the broadest model menu (75+ providers). Goose for a desktop GUI option, Linux Foundation governance, and general-purpose tasks beyond code.
What is Goose?
An open-source AI agent originally from Block, now governed by the Linux Foundation's Agentic AI Foundation at aaif-goose/goose. Written in Rust, ships as desktop app, CLI, and API, works with 15+ providers, and connects to 70+ MCP extensions. General-purpose, not code-only.
How many GitHub stars does OpenCode have?
172,198 stars (MIT), the most-starred open-source coding agent, ahead of Gemini CLI (105k) and Codex (90k).
Are they free?
Yes. OpenCode is MIT, Goose is Apache-2.0. You pay only for the model API you connect.
Can I use my Claude subscription?
Not a Claude Pro/Max subscription with OpenCode: Anthropic prohibits it for third-party tools, so OpenCode needs a paid Anthropic API key. Goose can reuse Claude, ChatGPT, and Gemini subscriptions through ACP.
Related comparisons
OpenCode vs Cursor
Open-source, model-agnostic terminal agent vs the closed IDE. Bring-your-own-key vs bundled.
OpenCode vs Cline
Two open-source agents: terminal-native OpenCode vs the VS Code extension Cline.
OpenCode vs Aider
Modern multi-provider TUI agent vs the original git-native pair programmer.
OpenCode vs Gemini CLI
Provider-agnostic open-source agent vs Google's free-tier Gemini terminal agent.
OpenCode vs Kilo Code
Terminal-first OpenCode vs the VS Code agent that merged Roo and Cline ideas.
OpenCode vs Claude Code
Open-source terminal agent vs Anthropic's Claude Code.
WarpGrep Boosts Any MCP Agent
WarpGrep v2 adds 2-3 points on SWE-bench Pro to every model tested. It runs as an MCP server inside OpenCode, Goose, and any tool that supports MCP. Free for 100k requests, then $1 per 1M. Better search means better context means better code.
