How do you run MiniMax in Claude Code?

Claude Code sends every request in Anthropic's Messages format. It reaches another server when ANTHROPIC_BASE_URL is set to that server's address.

MiniMax serves its models in that same format at https://api.minimax.io/anthropic. OpenRouter and Ollama's cloud answer in it too.

So MiniMax needs no translation in the middle, unlike GPT. What it needs is the right key, the model names and the window. How GPT differs is in running GPT in Claude Code.

MiniMax's quickest way is its setup wizard. It installs Claude Code if it is missing and writes ~/.claude/settings.json:

By hand, MiniMax first has you unset ANTHROPIC_AUTH_TOKEN and ANTHROPIC_BASE_URL. Exported in a shell, they override the settings file. Then it puts these in the env block of ~/.claude/settings.json:

• https://api.minimax.io/anthropic as the base URL, and your M Plan API key as ANTHROPIC_AUTH_TOKEN
• MiniMax-M3.1-Flash[1m] as the main model and as the Opus, Sonnet and Haiku models
• CLAUDE_CODE_AUTO_COMPACT_WINDOW at 524288, the 512K window MiniMax recommends
• CLAUDE_CODE_EFFORT_LEVEL at max

To check it, run /status and /model in a session. The base URL should read api.minimax.io/anthropic.

Every request in that session then goes to MiniMax. Claude models are out of reach until you change the file back.

That works for one person. On a team, it is one MiniMax key and one settings file per laptop. Why a key on every laptop is the expensive part is in keeping API keys off laptops.

MiniMax has two kinds of key, and they cannot stand in for each other.

If a subscriber's balance still goes down, the tool is using the pay-as-you-go key. MiniMax's fix is to put the M Plan key there instead.

An M Plan's text usage runs in a 5-hour window and a weekly window. Each starts with your first use and resets on its own. Explore gives three times Go's usage, and Build seven and a half times.

Pay-as-you-go, MiniMax-M3.1-Flash costs $0.15 per million input tokens and $0.60 per million output tokens. MiniMax-M3 costs $0.30 and $1.20 up to 512K input tokens, and twice that above.

No. MiniMax calls the M Plan its subscription for individuals. It is designed for one developer's interactive use, and MiniMax says to keep its key secure and not share it.

One subscription can run in several tools at once, but every tool draws on the same usage limits.

For production use, MiniMax recommends pay-as-you-go. A CI job belongs on a pay-as-you-go key, not on someone's M Plan.

For one developer who wants only MiniMax, MiniMax's own setup on their own M Plan is the simpler answer. A gateway adds nothing there.

Selan does not connect a MiniMax M Plan or a MiniMax key. It reaches MiniMax through an OpenRouter key or an Ollama Cloud plan, which an owner connects once.

Developers install the selan CLI, sign in as themselves and name the model:

The part before the first colon is the provider. Everything after it is that provider's own model id, passed through as it is. So openrouter:minimax/minimax-m3.1-flash reaches M3.1 Flash the same way.

MiniMax is not a model Selan lists, so nothing has to be added to Selan first. The model appears in Claude Code's /model picker for the session. The details are in Run any model.

A MiniMax turn runs on the provider its name points at, and on no other. If that account is out of quota, the request fails rather than moving to the other one.

Selan lists MiniMax M3 on Ollama's cloud with a 1,000,000-token window, and M2.7 with 200,000. Ollama answers only when you name one of its models. A request that names no model never lands on it.

How Ollama itself runs in Claude Code is in using Ollama with Claude Code. All of this is on Providers and models. When OpenRouter alone is the simpler answer is in OpenRouter and Selan compared.

• One usage record. MiniMax turns land beside Claude turns, per person and per repository. Usage per developer shows what an owner sees.
• One spend limit. An OpenRouter turn is metered, so it counts against a member's limit. A turn on an Ollama plan does not. See spend limits.
• No OpenRouter or Ollama key on a laptop. Keys stay on the gateway. A developer holds a Selan session you can revoke.

• M3.1 Flash on Ollama. Ollama's cloud listed minimax-m3.1-flash on 9 November 2026, but Selan's Ollama list did not yet include it. Through Selan, M3.1 Flash runs on OpenRouter.
• Thinking and spend. MiniMax-M3.1-Flash always thinks, and a setting that turns thinking off has no effect on it. On OpenRouter, the thinking is metered too.
• The window. MiniMax recommends a 512K window for everyday coding, even on a 1M model. In its own comparison, results were similar and the cost was about 20% lower.

Free for up to three seats, and the Free plan needs no card.