Manus Flex

​

What it is

Manus Flex lets you run Manus with your own API key from a supported inference provider. You choose the provider, the model, and the effort level. Manus still plans the work, uses tools, and runs tasks in the same projects, automations, and execution environments. Flex is for users who already have an account, credits, or a contract with an inference provider. With Flex, you can use that provider relationship inside Manus.

---

​

How to use it

​

Connect your provider

1. Open the mode menu in the top-left corner of the workspace. The menu shows your current mode, for example Manus 2.0.
2. Select Manus Flex.
3. In Provider, select your inference provider.
4. In API key, paste your API key.
5. Read the confirmation message and confirm. The message explains which costs your provider bills and which costs use Manus credits.

​

Choose your model and effort level

1. Open Manage your Flex models from the mode menu. (Exact entry point unverified.)
2. In Primary model, select the model that runs your main work. The list depends on your provider.
3. In Effort, select a level. A higher level can improve quality, but it is slower and uses more of your provider’s usage.
4. Optional: turn on Use lightweight model for side tasks to send smaller tasks to a lighter model. (Behavior unverified.)

When Flex is active, the mode menu shows Flex and your model, for example Manus Flex · [model name].

​

Change your API key or provider

1. Open Manage your Flex models.
2. Select Change API key.
3. Enter the new key, or select a different provider and enter its key.

​

Switch back to standard Manus

1. Open the mode menu in the top-left corner.
2. Select a standard Manus mode, for example Manus 2.0 Max, Manus 2.0, or Manus 2.0 Lite.

Your new tasks then use Manus models and Manus credits again.

---

​

What you need to know

​

Supported providers

Manus Flex works with these providers:

• OpenAI
• Anthropic
• Google AI Studio
• xAI
• Meta
• OpenRouter
• Fireworks
• Modal

OpenRouter, Fireworks, and Modal are the first partners in the Flex Inference Partner Program. Through them, you can use models from many model families.

​

Who pays for what

Flex splits the cost of a task into two parts:

​

What stays the same

• Your workspace: Projects, files, automations, and connectors work as before.
• The agent: Manus still plans the task, calls tools, and runs work in its execution environments.
• Built-in models: Video, image, audio, and 3D generation still use the Manus built-in models, not your provider.

​

What depends on your provider

• Model list: The models in Primary model come from your provider. Different providers show different models.
• Rate limits and quotas: Your provider’s limits apply. If your key reaches a limit or runs out of balance, tasks can stop or fail. (Error behavior unverified.)
• Quality and speed: Results depend on the model and effort level you choose. Manus does not guarantee third-party model performance.

---

​

Tips

• Use a dedicated API key. Create a separate key for Manus. Then you can track usage and revoke the key without affecting your other apps.
• Set a spending limit with your provider. Agent tasks can run many steps. A limit in your provider’s dashboard protects you from unexpected costs.
• Start with a medium effort level. Compare the quality and cost on a few tasks, then raise or lower the effort level.
• Use OpenRouter to compare models. One OpenRouter key gives you access to models from several model families, so you can test them without new keys.
• Test on a real task first. Run one of your usual tasks with Flex and compare it with a standard Manus mode before you switch your automations.
• Switch modes for each task. Use Flex for tasks where your own model fits best, and a standard Manus mode for everything else.

---

​

FAQ

---

Further reading: Introducing Manus Flex