Claude Code + Minimax 2.7: Unlimited AI Coding on a Budget

Run Claude Code against Minimax 2.7's token plan and you get near-Opus-level coding at the cost of a flat subscription.

## What the token plan actually gives you The Plus plan unlocks 4,500 model requests every 5 hours; the Max plan lifts that to 15,000 requests every 5 hours. Because the counter resets on a rolling window, the creator treats it as effectively unlimited for overnight runs: "just run the tap, let it flow, let it develop, let it work."

## Why pair it with Claude Code Claude Code runs locally, understands your repo, and works headlessly — so the creator schedules builds overnight on a VPS that "runs 24 hours for me." Waking up to a finished build is framed as the normal workflow, not a stretch goal. The boxminingai.com site was built this way and the creator says "MiniMax actually performed very well for the website."

## The config change Edit settings.json and drop in three env vars: your Minimax token-plan API key (the one "that resets"), the ANTHROPIC_AUTH_TOKEN, and the base URL https://api.minimax.io/anthropic. International users must use the international portal — minimax.com is the slow China endpoint. If you paste into an existing JSON, copy the inner object and add a trailing comma or the file will break.

## Cost reality check On a token plan you're paying a fixed subscription, not per-token Anthropic pricing, so heavy overnight sessions don't rack up extra bills. Off-plan, Minimax 2.7 is still cheaper than Claude Sonnet/Opus, and the creator claims output is "near Opus level intelligence for programming."

## Caveats the creator flagged - Don't paste a fresh top-level object into a non-empty settings.json without the comma fix. - The China base URL is slow for non-Chinese users — stick to api.minimax.io. - Model quality is good, not identical to Opus; review the diff before merging.