This guide walks through the complete process for using Xiaomi MiMo free token credits with Claude Code, helping developers handle quota applications, API key creation, Claude Code installation, and model provider switching in a practical way. Keywords: MiMo, Claude Code, cc-switch.
Technical specifications are summarized here
| Parameter | Details |
|---|---|
| Topic | Connecting the Xiaomi MiMo API to Claude Code |
| Runtime | Node.js 18+ |
| Integration Protocol | HTTP API / CLI toolchain |
| Client Tools | Claude Code, cc-switch |
| Core Dependencies | npm, @anthropic-ai/claude-code, cc-switch |
| Campaign Period | 2026-04-28 to 2026-05-28 |
| Target Users | Individual developers, teams, enterprises |
| GitHub Stars | Not provided in the source |
This workflow guide solves the problem of low-cost access to AI coding assistants
During the campaign period, Xiaomi MiMo offers large-scale free token credits. The real value is not just “sign up and get credits,” but that these credits can be connected directly to coding agent tools such as Claude Code and Cursor. For developers who need to call large models continuously, this can significantly reduce experimentation and development costs.
The original information is scattered across the campaign page, the platform console, and third-party switching tools. The actual pain points are clear: the application entry points are fragmented, model switching is not intuitive, and CLI configuration can easily fail because of environment dependencies or incorrect model names. This article compresses all of that into one executable path.
AI Visual Insight: The image shows the main landing page and campaign entry for the article topic. It highlights the core messaging around “Xiaomi MiMo Free Tokens” and the Claude Code integration scenario. This is a typical campaign onboarding screenshot used to confirm that developers can start the application flow from the event page.
You should first confirm the campaign window and eligibility scope
The campaign runs from 00:00 on April 28, 2026 to 00:00 on May 28, 2026, Beijing time, for a total of 30 days, while supplies last. The participation threshold is low. Global users can apply, including individuals, teams, and enterprise users.
It is best to visit two entry points first: the campaign page for applying for free credits, and the platform page for registration, key creation, and ongoing API management. This dual-entry structure is common. The first entry drives applications, while the second provides console capabilities.
# Campaign application entry
https://100t.xiaomimimo.com
# Platform registration entry (includes trial credits)
https://platform.xiaomimimo.com?ref=274DCEThese two URLs serve different purposes: one for applying, and one for using the platform console.
After applying for free credits, you should immediately verify that the quota appears in the console
After you submit the application form, the system will notify you by email. You must register on the platform using the same email address you provided in the application. Otherwise, even if your application is approved, you will not be able to see the quota or API capabilities in the console.
AI Visual Insight: The image shows the MiMo API open platform homepage or the campaign application form area. It indicates that the platform is positioned as an inference service entry point for global developers and supports integration with toolchains such as Claude Code, Cursor, and OpenClaw.
AI Visual Insight: The image shows the campaign application form interface. Technically, it suggests that quota approval depends on structured information submission rather than instant activation. That means developers need to pay attention to email verification and manual or semi-automated review steps.
AI Visual Insight: The image shows an email notification or registration confirmation interface. It indicates that quota distribution is tied to the account system, and the email address serves as the single identity anchor across application, registration, and credit redemption.
After registration succeeds, visit the following page to review your plan and quota: https://platform.xiaomimimo.com/console/plan-manage. In the source example, the account received 700 million free tokens, which suggests that actual allocations may vary dynamically according to campaign rules.
AI Visual Insight: The image shows the quota or plan management page in the console. The key technical point is that token quotas are centrally managed at the platform account level, and subsequent API calls, model testing, and usage statistics all depend on the data confirmed on this page.
The prerequisites for installing Claude Code on Windows are straightforward
Claude Code is essentially a JavaScript/TypeScript-based CLI tool, so your local machine must have a Node.js runtime installed first. The minimum recommended version is Node.js 18+, or installation and execution may fail.
AI Visual Insight: The image shows the Node.js download page. The technical takeaway is that Claude Code is not a standalone binary. It is distributed through the Node.js ecosystem, so environment preparation comes before model configuration.
node --version # Verify that Node.js is installed successfully
npm --version # Verify that npm is availableThe goal here is to confirm that the CLI runtime is fully available.
When installing Claude Code, users in mainland China may want to switch the npm registry first, then perform the global installation and version check.
# Set a local mirror registry to speed up package downloads
npm config set registry https://registry.npmmirror.com
# Install Claude Code
npm install -g @anthropic-ai/claude-code
# Check the installed version
claude --version
# First launch
claudeThese commands complete the Claude Code installation, verification, and first launch.
AI Visual Insight: The image shows the version check results in a terminal, confirming that Node.js is installed correctly and that the CLI tool can now be managed and executed through npm.
AI Visual Insight: The image shows the Claude Code installation or startup interface. It reflects that the workflow has moved from the dependency layer to the application layer. The next focus is no longer the environment, but provider integration.
Using cc-switch minimizes the complexity of switching between model providers
The value of cc-switch is not simply that it can configure models. Its real value is that it avoids manual configuration file edits. It provides a desktop UI for managing multiple CLI providers in one place and includes presets, SQLite-backed storage, and atomic writes. It is especially useful for developers who frequently test different models.
Download it here: https://github.com/farion1231/cc-switch/releases. After installation, go to the MiMo console and generate a dedicated API key for provider configuration.
AI Visual Insight: The image shows the cc-switch releases page or installer interface. It makes clear that cc-switch is a standalone desktop application responsible for multi-CLI configuration management, not the model inference service itself.
AI Visual Insight: The image shows where to obtain an API key in the MiMo console. The key detail is that integration authentication uses a key-based mechanism. Developers must generate and securely store keys on the platform side and cannot reuse application form data directly.
The model name is the easiest field to get wrong when adding a provider
After adding a Xiaomi MiMo provider in cc-switch, you need to fill in the API endpoint, API key, and model name. The source article shows a 400 error: Not supported model MiMo-V2.5-Pro. This indicates that the platform does not accept that model identifier. The root cause is a parameter mismatch, not a network or authentication failure.
Xiaomi MiMo check failed: HTTP 400:
{
"error": {
"code": "400",
"message": "Not supported model MiMo-V2.5-Pro",
"param": "Param Incorrect"
}
}This error strongly suggests that you should first verify the model name against the current list of supported models on the platform.
AI Visual Insight: The image shows the provider configuration form inside cc-switch. It demonstrates how the tool collects key parameters such as endpoint, model, and key through a graphical interface, reducing the risk of manually editing JSON or config files.
AI Visual Insight: The image shows an intermediate step in filling in provider parameters. It indicates that the model onboarding process includes field validation, and any issue with the model name, authentication header, or API base URL will surface during validation.
AI Visual Insight: The image shows the result page for a failed model check. Technically, it means cc-switch successfully sent the request, but the server rejected it with a 400 parameter error. Your troubleshooting should therefore focus on the model ID rather than the local network.
If you run into the same issue, use this troubleshooting order: first check the latest model ID in the platform documentation, then verify that the base URL matches the provider template, and finally confirm that the API key belongs to the correct account and plan.
You can only validate the full integration after testing end to end in Claude Code
Once cc-switch passes its checks, switch back to Claude Code and run a natural language task. Only then can you confirm that all four layers—quota, key, model, and CLI—are working together.
The source example asks Claude Code to generate a Markdown table of AI news items.
Help me compile the latest 20 AI news items in China for May 2026, then give me a summary, and output the result as a Markdown table in the current directory.This prompt is used to verify that the model can respond reliably in a real coding assistant workflow.
AI Visual Insight: The image shows the model test result page in cc-switch, indicating that the provider configuration has reached a callable state and has completed at least a connectivity test and a basic response validation.
AI Visual Insight: The image shows a further confirmation interface after model testing. It emphasizes that “connection successful” and “response correct” are two different stages, and the latter better demonstrates that the parameters are fully aligned.
AI Visual Insight: The image shows Claude Code in actual use. Its technical significance is that the upstream model provider has been successfully adopted by Claude Code, allowing the user to drive coding or content tasks directly through natural language.
AI Visual Insight: The image shows Claude Code executing the task prompt, demonstrating that this integration supports not only simple connectivity tests but also structured output workflows common in daily development.
The final conclusion is that this combination works well for low-cost AI coding experiments
If your goal is to experience agent-style programming at a lower cost, Xiaomi MiMo free credits plus Claude Code is a very practical path. The key is not the installation commands themselves, but the closed-loop workflow that connects credit application, account registration, key generation, model switching, and CLI validation.
For teams, the graphical provider-switching capabilities in cc-switch are especially valuable. They reduce the probability of environment drift and accidental configuration changes, and they make multi-model evaluation more efficient.
FAQ
1. Why do I have free credits but still cannot call the model from Claude Code?
Usually, the problem is not the quota itself. Instead, the account registration on the platform may be incomplete, the API key may not have been created, or the model provider may not have been configured correctly in cc-switch. First verify the available quota in the console, then verify the key and model name.
2. What should I do if I get an HTTP 400 Not supported model error?
In most cases, this means the model ID is incorrect. Check the latest MiMo platform documentation or the supported model list in the console first, then make sure the model name, API base URL, and provider template all match before testing again.
3. Why is Node.js 18+ required?
Claude Code depends on Node.js at runtime. If the version is too old, installation can fail, dependencies may become incompatible, or the CLI may not launch at all. That is why Node.js 18+ is the minimum practical prerequisite.
Core summary: This article reconstructs the full workflow for the Xiaomi MiMo large-scale token campaign and Claude Code integration. It covers eligibility, quota verification, Node.js and Claude Code installation, cc-switch configuration, API key creation, and common error troubleshooting. It is especially suitable for developers who want low-cost access to large-model coding assistants.