MiniMax's official OpenClaw page provides an M Plan OAuth wizard: select the MiniMax provider and MiniMax Global — OAuth (minimax.io), then set the exact model MiniMax-M3.1-Flash as the default; the page does not publish an OpenClaw provider ID, API Base URL, or API request verification command.
Suitable tasks: Using MiniMax M-series models in a local OpenClaw installation and connecting channels such as Telegram, WhatsApp, Discord, and iMessage.
Unsuitable tasks: Situations that require manually entering an API endpoint, environment variable, or provider routing field and cannot accept an integration flow in which the page does not publish these parameters.
Supported model version: MiniMax-M3.1-Flash. The page requires setting it as the default model in the model picker.
Supported client, Agent, or API: OpenClaw; the page describes it as a locally running personal AI assistant.
Recommended reasoning mode and parameters: Thinking is always enabled; do not use /think off. Other reasoning parameters are not stated.
The following command and selections only record the official wizard; they are not executed locally:
curl -fsSL https://openclaw.ai/install.sh | bash| Configuration item | Official page value |
|---|---|
| Model provider | MiniMax |
| Authentication method | MiniMax Global — OAuth (minimax.io) |
| Default model | MiniMax-M3.1-Flash |
| Prerequisite credential | MiniMax Subscription Key with a valid M Plan subscription |
| API Base URL | Not stated |
| OpenClaw provider ID / model routing field | Not stated |
| API key environment variable or manual entry field | Not stated; the page only says that a Subscription Key is required and directs the user to browser-based OAuth |
Confirm that you have a MiniMax Subscription Key associated with a valid M Plan subscription, as well as a macOS, Linux, or Windows + WSL environment with terminal access.
Run the installation command above, select Yes when OpenClaw setup starts.
Select QuickStart, then set the model provider to MiniMax.
Set the authentication method to MiniMax Global — OAuth (minimax.io), sign in to or register a MiniMax API Platform account in the browser, and click Authorize.
Return to the terminal, select the exact model MiniMax-M3.1-Flash in the model picker, and press Enter to set it as the default model.
Select the messaging channels to connect, complete the configuration as prompted, and confirm the installation.
Select additional skills as needed; configure the other API keys required by the tools, or select Skip for now if they are not currently needed.
Select Open the Web UI to open the OpenClaw control panel.
The results verifiable from the page are that the provider, OAuth method, and default model have been selected, and that the Web UI can be opened. The page does not provide an independent command for sending a test request, checking the model response, inspecting the API route, or checking health status, so those verification parameters are not stated.
The page describes OpenClaw as a locally running personal AI assistant that can be controlled through messaging platforms.
The prerequisites are a MiniMax Subscription Key associated with a valid M Plan subscription, plus a macOS, Linux, or Windows + WSL terminal environment.
The wizard explicitly provides the MiniMax provider, the MiniMax Global — OAuth (minimax.io) authentication method, and MiniMax-M3.1-Flash as the default model.
The page lists Telegram, WhatsApp, Discord, iMessage, and other connectable channels, but does not describe the limitations or configuration fields for each channel.
Additional API keys are optional and are used for tools that OpenClaw can call; when they are not needed, Skip for now can be selected.
The page explicitly states that thinking is always enabled for MiniMax-M3.1-Flash and cannot be disabled; do not use /think off.
The page body does not publish an OpenClaw provider ID, model routing configuration, Base URL, API endpoint, environment variable name, M Plan quota, rate limits, timeout, or model fallback policy.
The entry screenshot shows OpenClaw 2026.3.23-2; the body does not state that this version is required, so no version requirement can be added on that basis.
The OAuth flow requires signing in to or registering a MiniMax API Platform account in the browser and authorizing it; the page does not state whether other key injection methods are available.
Subscription Key is listed as a prerequisite, but the page does not say whether it is pasted directly into OpenClaw, nor does it provide a key variable name or configuration file location.
The page does not provide an API Base URL or provider routing field, so manual JSON, environment variables, or an OpenAI/Anthropic-compatible endpoint cannot be inferred from it.
OpenClaw's entry-point security notice says that the project is still in beta and uses a personal-agent and trusted-operator boundary by default; access controls and security hardening should be completed before enabling tools or exposing it to the internet.
Thinking cannot be disabled for MiniMax-M3.1-Flash; this article does not provide a configuration for disabling thinking or use /think off as a verification step.
The page does not state the required OpenClaw version, messaging-channel quotas, model rate limits, timeouts, costs, context window, or failure-retry rules.
On 2026-10-21, the official page was opened and read through the Tabbit exclusive task M31 prompt10 openclaw, and its body, code elements, and entry-point security notice were checked; the extracted source evidence is saved at /tmp/m31-openclaw-evidence.txt. OpenClaw was not installed, MiniMax was not accessed, OAuth was not authorized, no key was entered, and no user configuration was changed.
MiniMax M3.1 Flash