AINews wire, 3 h ago

OpenCode: a free open-source AI coding tool with a desktop app and model choice

Reported by XDA Developers

Image: XDA Developers

A reviewer at XDA Developers says OpenCode, a free open-source AI coding tool, persuaded them to skip buying another paid coding subscription. They already paid for Claude Code, Codex and Google Antigravity, and were close to buying Cursor when they found OpenCode.

At first the reviewer assumed OpenCode was only a terminal tool. They found it has a standalone desktop app with a visual interface for managing projects, reviewing changes and tracking what AI agents are doing. The app also includes a built-in browser, so results can be checked without switching applications.

Model choice is the main draw. According to the review, users can connect several AI providers, use a set of free models, or run local models through Ollama. Cheap or free models can handle small jobs such as minor styling fixes or simple components, while premium models can be used for harder work.

OpenCode includes Plan and Build agents. Plan maps out a project without changing files, and Build then carries out the changes. The tool can edit files, run terminal commands and handle multi-step development tasks. Users can create custom agents, set permissions and add features through MCP servers.

To test it, the reviewer asked OpenCode to build a complete travel website from scratch. Free models set up the foundation, generated reusable components and handled basic styling. Premium models were used to refine the interface, add complex interactions and fix stubborn problems. The reviewer said switching between models did not disrupt the workflow, and the project never had to be exported to another application.

The first result was not perfect. After several rounds of changes, the reviewer had a responsive, polished travel site with pages, filters and a booking interface. The reviewer describes OpenCode as an open-source rival to Claude Code and Codex, and says it has earned a place in their workflow, although they are not dropping their current tools yet.

The article is one reviewer's personal experience, not a benchmark. Results will depend on which models a user chooses and on the size of the project. The report does not measure speed or output quality against rival tools. It also does not give detailed pricing for the premium models, the exact number of free models, or the full licence terms. Readers should check the project's own pages for those details.

Enjoying this? Get the AI brief

Every Thursday: AI that matters for your home.

✓ Free · seven editions · one click to leave

Topics
Story continues on XDA DevelopersAdd us as a preferred source on GoogleSee our stories first in Google’s Top Stories.
Get alerts on big newsOnly the big news, at most once a day. No email, no account.

Original reporting by the sources above. Summary by smart-home.space.