OpenCode has the profile of a coding agent that is being treated like developer infrastructure. ToolVitals tracks 161,632 GitHub stars, a 96 ToolVitals score, a 95 health score, and a 100 shipping score. The sharpest signal is the release cadence: 13 release events in 30 days and 30 GitHub releases in 90 days.

The website positions OpenCode as an open source AI coding agent for the terminal, IDE, and desktop. It also says the product supports 75+ LLM providers through Models.dev, includes GitHub Copilot and ChatGPT Plus or Pro login paths, and has a desktop app in beta for macOS, Windows, and Linux. That matters because the release notes are not only cosmetic updates. They show work on the surfaces where coding agents usually break: shells, editors, language servers, model quirks, permissions, and session state.

The late April release train is unusually dense. v1.14.21 added pull diagnostics for LSP servers, Roslyn Language Server support for C#, better session compaction, and fixes for bare Git repos and worktrees. v1.14.24 added experimental HTTP API endpoints for MCP server status, file listing, file reads, and project file status. v1.14.25 tightened permission config behavior and LSP permission prompts. v1.14.27 added a configurable default shell for terminals and agent shell commands. v1.14.28 fixed an upgrade path for Bun installs.

That pattern suggests a team betting on integration depth, not just chat UX. OpenCode is trying to sit inside real development loops: local shells, project worktrees, LSP diagnostics, model-provider differences, desktop state, TUI state, and MCP-adjacent API access. Boring fixes count here. A coding agent that mishandles .npmrc, loses working directories after shell startup, or formats reasoning output incorrectly will feel broken no matter how good the model is.

OpenCode is OSI-approved OSS under the MIT license according to the ToolVitals openness signal. That label is stronger than source-available or fair-code. It means ToolVitals can call it open source without watering the wording down.

What ToolVitals cannot infer

ToolVitals can see public maintenance signals. It can see stars, release events, scores, website availability, SSL status, and public release notes. It cannot prove code quality, security posture, enterprise readiness, user satisfaction, revenue, or whether OpenCode works well on your specific stack.

The GitHub repository page also showed activity beyond the supplied metrics, including a recent commit and public repository positioning, but the ToolVitals counts remain the source of truth for this post. Treat the 161,632 stars, 13 release events in 30 days, 30 releases in 90 days, and 96 ToolVitals score as the canonical numbers.

Comparisons

Against nearby developer tools, OpenCode is not the loudest by every metric. OpenClaw shows 372,593 stars and 36 release events in 30 days. LangChain shows 136,938 stars and 20 release events in 30 days. Gemini CLI shows 104,156 stars and 10 release events in 30 days.

OpenCode sits in the middle of that pack on release count, but its 161,632 stars and 100 shipping score put it well above normal project-maintenance noise. n8n has a higher hot score at 240.0 and 45 release events in 30 days, but ToolVitals classifies n8n as fair-code, not OSI-approved open source. That distinction matters if your team cares about license rights as much as activity.

Recommendation

If your team wants a coding agent that can run across terminal, IDE, and desktop workflows, evaluate OpenCode because its recent releases target the messy integration points that decide whether agents survive daily use. Start with a small repo that uses real LSP diagnostics, shell startup files, and at least one non-default model provider. That is where OpenCode’s shipping pattern should show up fastest, either as useful polish or as bugs you can report against a very active MIT-licensed project.

Sources