It's not a fork of OpenCode. Instead, it implements the OpenCode protocol and just runs `opencode attach` to the server that converts API calls to the underlying agents.
We build this to scratch our itch of being able to rapidly switch between coding agents based on the task at hand. For example, we find that:
- Claude Code is the best executor & fast iterator - Codex (high) is the best for complex or long-running tasks - OpenCode for fine-tuned, do-exactly-as-I-say edits
I personally believe that harnesses matter almost as much as the models in 2026. OpenCode lets you swap out models already, but the CC & Codex harnesses + system prompts make a big difference in practice.
Under the hood, this is all powered by our Sandbox Agent SDK:
- Sandbox Agent SDK provides a universal HTTP API for controlling Claude Code, Codex, and Amp - Sandbox Agent SDK exposes an OpenCode-compatible endpoint so OpenCode can talk to any agent - OpenCode connects to Sandbox Agent SDK via attach
I want to emphasize: the Anomaly folks are doing awesome work with OpenCode agent + Zen + Black. I use OC regularly alongside CC & Codex depending on the task. Gigacode is only possible because OpenCode is insanely flexible, hackable, and well documented.
Give it a try:
$ curl -fsSL https://releases.rivet.dev/sandbox-agent/latest/gigacode-ins... | sh
Check out the project, architecture, and other install options:
https://github.com/rivet-dev/sandbox-agent/tree/main/gigacod...
