Technology RadarTechnology Radar
This item was not updated in last three versions of the Radar. Should it have appeared in one of the more recent editions, there is a good chance it remains pertinent. However, if the item dates back further, its relevance may have diminished and our current evaluation could vary. Regrettably, our capacity to consistently revisit items from past Radar editions is limited.
Trial

OpenCode is an open-source AI coding agent for the terminal, built by the team behind SST (Serverless Stack). It has grown to 120K+ GitHub stars and 5M monthly users since its June 2025 launch — making it one of the fastest-growing developer tools in recent memory.

Architecture Deep Dive → OpenCode Architecture Breakdown — client/server Go architecture, native LSP error feedback loop, multi-provider routing via Models.dev, and how it compares to Claude Code and Codex CLI.

Why It Matters for Engineering Managers

OpenCode has emerged as the open-source alternative to Anthropic's Claude Code and OpenAI's Codex CLI. Its growth trajectory (surpassing 1.1 million total downloads within months) suggests it is meeting real developer demand. For teams evaluating AI coding agents, it belongs in your assessment.

The key differentiators relevant to a software organisation:

  • No vendor lock-in: Supports 75+ LLM providers through Models.dev. Switch between Claude, GPT-5-Codex, Gemini, and local models without changing your workflow.
  • Privacy architecture: OpenCode stores no code or context data itself — all data goes directly to whichever LLM provider you configure. For teams with IP concerns, this is meaningful.
  • Open source (MIT): Full auditability. Security teams can review the codebase; no black box in your development workflow.
  • LSP integration: When the agent makes a code edit that introduces a syntax error, the Language Server Protocol error is automatically fed back to the model — closing the loop without human intervention.

Built-in Agents

OpenCode ships with two agents you can switch between:

  • Build (default): Full read/write/shell access — the standard development agent
  • Plan: Read-only mode for analysis, code exploration, and planning without risk of unintended changes. Good for onboarding agents to a new codebase.

Architecture

OpenCode uses a client/server design: the agent server runs on your machine, and the terminal UI (TUI) is just one possible client. This architecture opens the door to driving it from other interfaces (IDE extensions, mobile apps, web UI) while keeping the computation and file access local.

Installation

# Quick install
curl -fsSL https://opencode.ai/install | bash

# Or via npm
npm install -g opencode-ai

# Or Homebrew
brew install opencode

Compared to Claude Code

Both are terminal-first agentic coding tools. The key trade-off:

OpenCode Claude Code
Model support 75+ providers Claude (primarily)
License MIT (open source) Proprietary
Data handling Direct to your LLM provider Anthropic API
LSP error feedback ✅ Native ❌ Not built-in
Adoption 5M users/month #1 tool by survey (Feb 2026)

Key Characteristics

Property Value
Interface Terminal (TUI) + IDE extensions
License MIT
Model support 75+ providers (Claude, GPT, Gemini, local)
Provider SST / opencode-ai (community)
Website opencode.ai
GitHub sst/opencode

Further Reading