How to Set Up OpenClaw with DeepSeek on Agent 37

Do not index

If you want to run OpenClaw with DeepSeek, this guide walks you through the full setup inside Agent 37. You will create a DeepSeek API key, open your hosted OpenClaw terminal, and configure DeepSeek as the active model provider for your instance.

This page is based on the video walkthrough below and turns it into a written step-by-step guide you can follow at your own pace.

Video walkthrough

💡

Before you begin, you need an active Agent 37 account, a running OpenClaw instance in Agent 37, a DeepSeek API account, and a valid DeepSeek API key.

Why use DeepSeek with OpenClaw

DeepSeek is a useful option when you want a provider that is API-compatible with OpenAI-style tooling while still giving you DeepSeek-specific model choices. If you already run OpenClaw in Agent 37, switching to DeepSeek is mostly a configuration change in the hosted terminal.

Step 1: Create your DeepSeek API key

Go to the DeepSeek API docs and open your DeepSeek platform account.

Create or copy an API key for the account you want to use with OpenClaw.

Store the key somewhere safe so you can paste it into the Agent 37 terminal.

Keep this key private. Anyone with the key can consume usage on your DeepSeek account.

Step 2: Open your Agent 37 instance

Go to the Agent 37 dashboard.

Open the OpenClaw instance you want to configure.

Click Terminal.

You will do the full setup from the hosted Agent 37 terminal. You do not need to configure DeepSeek on your local machine.

Step 3: Start the OpenClaw configuration flow

In Terminal, run:

openclaw configure

When prompted, choose:

Where will the gateway run? Choose Local

Select Section to Configure? Choose Model

Model / Auth Provider Choose DeepSeek

Step 4: Add your DeepSeek API key

After selecting the provider:

Choose the DeepSeek API key authentication option

Paste the API key you created in Step 1

Continue to the model picker

OpenClaw will save the credential to your instance configuration.

Step 5: Select a DeepSeek model

Choose the DeepSeek model you want OpenClaw to use as the active model for this instance.

Common choices include:

deepseek-v4-flash
deepseek-v4-pro

According to the official DeepSeek API docs, deepseek-v4-flash is the fast general-purpose model and deepseek-v4-pro is the stronger reasoning model. The older deepseek-chat and deepseek-reasoner names were retired in July 2026, so avoid them in new setups.

Step 6: Verify the model is active

Run:

openclaw models list

You should see the selected DeepSeek model in the output.

If you want to switch models later, rerun openclaw configure or use the exact model identifier shown by OpenClaw in the model management commands.

What this changes in Agent 37

Once setup is complete:

your OpenClaw instance uses DeepSeek as the configured provider

your agent can use the selected DeepSeek model for prompts and workflows

you can keep using the same Agent 37 channels, tools, and hosted environment

Common issues

If the setup does not work on the first try, check these first:

the API key was copied correctly

you selected DeepSeek in openclaw configure

the model name you picked is available in your DeepSeek account

the configuration flow finished without interruption

your DeepSeek account has available balance if the API rejects requests

If needed, rerun:

openclaw configure

Which DeepSeek model should you choose

Start with deepseek-v4-flash if you want a fast general-purpose model for everyday agent tasks. It is the second most-used model across real OpenClaw traffic, and our ranking of the best models for OpenClaw explains where it beats the pricier options and where it does not.

Choose deepseek-v4-pro if your workflow benefits from stronger reasoning and more deliberate responses.

What to do next

Change the default model in OpenClaw as your workflow changes.

Set up OpenClaw with an OpenAI API key to compare provider flows.