← Back to research
·4 min read·opensource

OpenCode

OpenCode is an open-source AI coding agent with 101K GitHub stars, supporting 75+ LLM providers across terminal, desktop app, and IDE extensions.

Key takeaways

  • 101K GitHub stars makes it the most popular open-source coding agent by far
  • Desktop app (macOS/Windows/Linux), terminal CLI, and IDE extensions (VS Code, Cursor, Zed, Windsurf)
  • Model-agnostic: works with Claude, GPT, Gemini, local models, and 75+ providers

FAQ

What is OpenCode?

OpenCode is an open-source AI coding agent that runs in terminal, desktop, or IDE, supporting any LLM provider.

How is OpenCode different from Claude Code?

OpenCode is 100% open source and provider-agnostic, while Claude Code is Anthropic-only. OpenCode also has native LSP support and a client/server architecture.

Is OpenCode free?

Yes, OpenCode is free and open source. Users provide their own API keys or use the optional Zen service for optimized models.

Does OpenCode compete with Tembo?

OpenCode is a coding agent; Tembo orchestrates coding agents. They're complementary — Tembo can delegate tasks to OpenCode.

Project Overview

OpenCode is the most popular open-source AI coding agent, with over 101,000 GitHub stars, 650+ contributors, and 2.5 million monthly developers.[1] Built by the creators of terminal.shop, it provides a provider-agnostic alternative to Claude Code and Codex.[2]

The project is available as a desktop app (macOS, Windows, Linux), terminal CLI, and IDE extensions — all backed by a client/server architecture that enables unique capabilities like driving the agent remotely from a mobile app.[3]

What It Does

OpenCode is a coding agent that can:[3]

  • Read and write code in any language
  • Execute shell commands
  • Navigate and understand codebases via LSP integration
  • Work with any LLM provider (75+ supported)

Built-in agents:

AgentPurpose
buildFull-access agent for development (default)
planRead-only agent for analysis and exploration
generalSubagent for complex searches and multi-step tasks

Technical Architecture

Platforms:[1]

SurfaceAvailability
Desktop AppmacOS (Apple Silicon & Intel), Windows, Linux (.deb, .rpm) — Beta
Terminal CLIAll platforms via npm, Homebrew, Scoop, etc.
IDE ExtensionsVS Code, Cursor, Zed, Windsurf, VSCodium
IntegrationsGitHub, GitLab

Model support: OpenAI, Anthropic, Google, local models via 75+ providers through Models.dev

Key differentiators from Claude Code:[3]

  • 100% open source (Apache 2.0)
  • Provider-agnostic (not locked to one model)
  • Native LSP support (automatic language server integration)
  • Client/server architecture (run locally, control remotely)
  • TUI-first design (built by neovim users)

Optional services:

  • Zen — Curated set of AI models optimized and benchmarked for coding agents
  • GitHub Copilot — Log in with GitHub to use existing Copilot subscription
  • ChatGPT Plus/Pro — Log in with OpenAI to use existing subscription

Strengths

  • Massive community — 101K stars, 650 contributors, 8,500+ commits [4][5]
  • Provider freedom — Use any model from any provider
  • Cross-platform desktop app — macOS, Windows, Linux in Beta
  • IDE everywhere — VS Code, Cursor, Zed, Windsurf, VSCodium
  • LSP integration — Automatic language intelligence out of the box
  • Privacy-first — No code or context stored on OpenCode servers
  • GitHub/GitLab native — Direct integration with major code forges
  • Active development — Fast-moving team with frequent releases
  • Enterprise-ready — Privacy-sensitive environment support

Weaknesses / Risks

  • No cloud orchestration — Local-first means no background agent support
  • Requires API keys — Users manage their own model access
  • Community support — No enterprise SLAs or dedicated support
  • Fragmented interfaces — Desktop, CLI, and IDE may have different capabilities
  • No enterprise integrations — No Jira, no signed commits, no compliance features

Competitive Landscape

Claude Code — Anthropic's terminal agent. Locked to Claude models but deeply optimized for them. OpenCode is the open alternative.

Codex CLI — OpenAI's equivalent. Similar positioning but OpenAI-only. OpenCode is provider-agnostic.

Cursor — IDE-integrated. Different form factor (editor vs. terminal). OpenCode is for developers who prefer CLI workflows.

Tembo — Agent orchestration layer. Complementary — Tembo can delegate tasks to OpenCode as one of many supported agents.

Ideal User

OpenCode is built for developers who:

  • Live in the terminal and prefer CLI tools
  • Want model flexibility (not locked to one provider)
  • Value open source and community-driven development
  • Need LSP integration without configuration
  • Want to control the agent from multiple interfaces

Pricing

Free and open source under Apache 2.0.[3]

Zen (optional) — Subscription for optimized model access; pricing not public.

Bottom Line

OpenCode is the clear community favorite among open-source coding agents. The 95K stars and massive contributor base signal real momentum and staying power.

The model-agnostic approach is a strategic advantage as the LLM landscape evolves — OpenCode users aren't locked in if a better model emerges. The client/server architecture is genuinely innovative, enabling use cases (like mobile control) that competitors can't match.

For teams needing enterprise features (signed commits, Jira integration, background agents), OpenCode isn't the answer. But for individual developers who want a powerful, flexible coding agent without vendor lock-in, it's the obvious choice.


Research by Ry Walker Research • methodology