arthur-albuquerque/claude_opencode_settings — explained in plain English
Analysis updated 2026-05-18
Run long Claude Code sessions without burning through Claude's usage limits by offloading code writing to cheaper worker models.
Set up a clear split where Claude handles diagnosis and review while OpenCode Go models handle the actual code edits.
Track two separate usage budgets, one for Claude and one for the OpenCode Go worker plan, so you never get surprised by a block.
| arthur-albuquerque/claude_opencode_settings | 0xhassaan/nn-from-scratch | 3ks/embedoc | |
|---|---|---|---|
| Stars | 0 | 0 | — |
| Language | Python | Python | Python |
| Last pushed | — | — | 2023-06-08 |
| Maintenance | — | — | Dormant |
| Setup difficulty | moderate | moderate | hard |
| Complexity | 3/5 | 4/5 | 1/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an active OpenCode Go subscription plus manually merging JSON config into ~/.config/opencode/opencode.json.
claude_opencode_settings is a configuration setup for people who use Claude Code alongside a separate OpenCode Go subscription. Instead of using Claude for every step of writing code, it turns Claude into a coordinator that handles the thinking work, diagnosing problems, breaking tasks down, reviewing code, and checking that things actually work, while handing off the actual typing of code to cheaper OpenCode Go models through the opencode command line tool, all from within the same Claude session. The point of this setup is to let long coding sessions run without running out of Claude's usage limits, because the bulk of token spending, the part where code actually gets written, happens on a flat rate worker plan instead of on Claude itself. The repository contains three files that install into your existing Claude Code and OpenCode setup. A CLAUDE.md file becomes the global system prompt and covers delegation rules, a table ranking which worker model to use for which kind of task, how to pace usage across two separate budgets, and a way to automatically resume long jobs later. A Python script checks how much budget remains on the OpenCode Go worker plan by sending a very small test request that does not use up any of the paid quota. A JSON file provides an example worker agent configuration that must be merged into your OpenCode settings, since without it OpenCode will silently refuse to let the worker model edit files. Setup involves installing the opencode command line tool with an active OpenCode Go subscription, having Node and Python 3 available, copying the CLAUDE.md and script files into the right folders, and merging in the worker agent configuration. Once that is done, every Claude Code session on the machine automatically follows the coordinator and worker split. The author notes a few limits to be aware of. The list of recommended worker models and their pricing reflects OpenCode Go's lineup as of July 2026 and will need updating as that catalog changes. The budget numbers assume a specific twelve dollar per five hour plan, so anyone on a different plan needs to adjust the figures themselves. The automatic resume feature only works while the Claude Code session and the machine it runs on stay awake, and the budget checking script can only tell you whether the worker plan is currently blocked, since OpenCode Go does not offer a way to check an exact percentage of usage remaining.
A settings package that turns Claude Code into a coordinator directing cheaper OpenCode Go models to write code, so long sessions use less of Claude's usage limit.
Mainly Python. The stack also includes Python, Claude Code, OpenCode CLI.
No license information is stated in the README.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.