Model providers

Using DeepSeek with OpenClaw

3 min read

Browse more in Model providers.

All model providers guides →

This guide walks you through configuring DeepSeek as a model provider in OpenClaw using the DeepSeek OpenAI-compatible API. You will wire up your DeepSeek API key, onboard the provider via the OpenClaw CLI, and understand which built-in DeepSeek models you can target.

By the end, your OpenClaw gateway will talk to DeepSeek using `deepseek/deepseek-chat` or `deepseek/deepseek-reasoner` as models.

Setup flow

Prerequisites

  • ✓An active DeepSeek account with an API key from https://platform.deepseek.com/api_keys.
  • ✓An OpenClaw installation with the `openclaw` CLI available in your shell.
  • ✓Access to the machine where your OpenClaw Gateway runs so you can set environment variables or edit `~/.openclaw/.env`.

Steps

  1. 1

    Get your DeepSeek API key

    Before you touch OpenClaw, you need a DeepSeek API key that the gateway can use for authentication. com`, and OpenClaw talks to it using that key.

    Log into DeepSeek and create an API key at the URL from the docs so you have a token ready for the next steps.

    text
    Get your API key at https://platform.deepseek.com/api_keys.
  2. 2

    Onboard DeepSeek interactively with the OpenClaw CLI

    Use the OpenClaw CLI to onboard DeepSeek so the gateway knows about the provider and can store your key. This interactive flow prompts you for the API key and configures `deepseek/deepseek-chat` as the default model, which is usually what you want for general chat-style agents.

    bash
    openclaw onboard --auth-choice deepseek-api-key
  3. 3

    Onboard DeepSeek non-interactively for automation

    If you script your infrastructure or CI, run the non-interactive onboarding command instead. This lets you pass `DEEPSEEK_API_KEY` from your environment, set the mode to `local`, skip health checks, and accept the risk flags without any prompts.

    It’s ideal for repeatable setups on servers.

    bash
    openclaw onboard --non-interactive \
  --mode local \
  --auth-choice deepseek-api-key \
  --deepseek-api-key "$DEEPSEEK_API_KEY" \
  --skip-health \
  --accept-risk
  4. 4

    Expose DEEPSEEK_API_KEY to the OpenClaw Gateway process

    When you run the OpenClaw Gateway as a daemon (launchd or systemd), the environment from your shell usually does not propagate automatically. shellEnv` configuration.

    If the daemon cannot see this variable, DeepSeek calls will fail even if onboarding worked.

    text
    If the Gateway runs as a daemon (launchd/systemd), make sure `DEEPSEEK_API_KEY` is available to that process (for example, in `~/.openclaw/.env` or via `env.shellEnv`).
  5. 5

    Choose the right DeepSeek model from the built-in catalog

    OpenClaw ships with a built-in catalog for DeepSeek so you can reference models by stable IDs. `deepseek/deepseek-chat` is the default non-thinking chat surface, while `deepseek/deepseek-reasoner` exposes the reasoning-enabled surface with a larger max output.

    Pick the one that matches your use case and configure your agents or calls accordingly.

    text
    Model ref: `deepseek/deepseek-chat`  | Name: DeepSeek Chat      | Context: 131,072 | Max output: 8,192  | Default model; DeepSeek V3.2 non-thinking surface
Model ref: `deepseek/deepseek-reasoner` | Name: DeepSeek Reasoner | Context: 131,072 | Max output: 65,536 | Reasoning-enabled V3.2 surface

Configuration

OptionDescriptionExample
DEEPSEEK_API_KEYDeepSeek API key used by OpenClaw to authenticate against the DeepSeek OpenAI-compatible API at https://api.deepseek.com.sk-deepseek-1234567890abcdef

Troubleshooting

OpenClaw Gateway cannot call DeepSeek when running under systemd, even though `openclaw onboard` succeeded.

The daemon process does not see `DEEPSEEK_API_KEY`, so authentication fails at runtime. shellEnv`, then restart the service so it picks up the new environment.

bash
If the Gateway runs as a daemon (launchd/systemd), make sure `DEEPSEEK_API_KEY` is available to that process (for example, in `~/.openclaw/.env` or via `env.shellEnv`).

The default model in OpenClaw is not the DeepSeek model you expect after onboarding.

The interactive DeepSeek onboarding sets `deepseek/deepseek-chat` as the default model. If you want the reasoning surface instead, explicitly configure `deepseek/deepseek-reasoner` in your agent or model configuration after onboarding so OpenClaw routes calls to the correct model.

bash
Model ref: `deepseek/deepseek-chat` (default)
Model ref: `deepseek/deepseek-reasoner`

Frequently asked questions

Powered by Mem0

Add persistent memory to OpenClaw

Official Mem0 plugin for OpenClaw keeps context across chats and tools. Smaller prompts, lower cost, better continuity for your agents.

Read the official Mem0 guideOpen Mem0

More in Model providers

Model providers

Using Alibaba Model Studio with OpenClaw

Model providers

Using BytePlus with OpenClaw

Model providers

Using fal with OpenClaw

Model providers

Using GLM (Zhipu AI) with OpenClaw

Model providers

Using Hugging Face with OpenClaw

Model providers

Using Inferrs with OpenClaw