OpenClaw with SEA-LION: Running Multilingual Personal AI Assistants

ByAISG | AI Products

OpenClaw is a personal AI assistant you run on your own devices. A growing user community configures their OpenClaw assistants to perform autonomous tasks. However, running OpenClaw with commercial API models can incur significant costs. SEA-LION provides free, capable models via the SEA-LION API, making it a good fit for:

  • Agent experiences that require Southeast Asia context and multilingual support
  • Always-on Heartbeat orchestration where calls are frequent and costs can add up

These steps assume you already have OpenClaw installed (https://docs.openclaw.ai/install) and set up on your machine.

A screenshot of a chat interface discussing activities people do at Gelora Bung Karno (GBK) in Jakarta, listing various events such as sports, concerts, jogging, and official government gatherings.

Add the SEA-LION model to OpenClaw

Add SEA-LION as a model provider in ~/.openclaw/openclaw.json. Replace <SEA_LION_API_KEY> with your API key from https://docs.sea-lion.ai/guides/inferencing/api.

Find the available SEA-LION models by querying the /models endpoint in https://docs.sea-lion.ai/guides/inferencing/api#step-1.-find-the-available-models. Assign the selected model in the OpenClaw configuration.

Alternatively, the recently added vLLM provider (as of February 2026) can be used to add the SEA-LION model.

{
  "models": {
    "providers": {
      "sealion": {
        "baseUrl": "https://api.sea-lion.ai/v1",
        "apiKey": "<SEA_LION_API_KEY>",
        "api": "openai-completions",
        "models": [
          { "id": "aisingapore/Qwen-SEA-LION-v4-32B-IT", "name": "Qwen-SEA-LION-v4-32B-IT" }
        ]
      }
    }
  }
}

Expose the SEA-LION model to the agents in agents.defaults.models:

{
  "agents": {
    "defaults": {
      "models": {
        "sealion/aisingapore/Qwen-SEA-LION-v4-32B-IT": { "alias": "Qwen-SEA-LION-v4-32B-IT" }
      }
    }
  }
}

Configure an OpenClaw Agent to use SEA-LION

Create the sea-lion agent and configure its identity in ~/.openclaw/workspace-sea-lion/IDENTITY.md. Refer to https://docs.openclaw.ai/cli/agents for the details.

openclaw agents add sea-lion

Customise the sample IDENTITY.md.

# IDENTITY.md - Who Am I?
**Name:** SEA-LION
**Creature:** AI Assistant
**Vibe:** Helpful, culturally-aware, multilingual
**Emoji:** 🌊
**Avatar:** https://playground.sea-lion.ai/images/favicon.ico

I am SEA-LION, your personal assistant specialized in Southeast Asian languages and culture.
I can communicate in multiple languages including Burmese, Indonesian, Malay, Filipino, Tamil, Thai, and Vietnamese.
I'm here to help you with cultural insights, language translation, travel advice, and any other questions you might have about Southeast Asia.

Update the agent to use the SEA-LION model via the web dashboard, or configure models.json in the agent configuration directory (e.g. ~/.openclaw/agents/sea-lion/agent/). Replace <SEA_LION_API_KEY> with the API key from https://docs.sea-lion.ai/guides/inferencing/api.

{
  "providers": {
    "sealion": {
      "baseUrl": "https://api.sea-lion.ai/v1",
      "apiKey": "<SEA_LION_API_KEY>",
      "api": "openai-completions",
      "models": [
        {
          "id": "aisingapore/Qwen-SEA-LION-v4-32B-IT",
          "name": "Qwen-SEA-LION-v4-32B-IT",
          "reasoning": false,
          "input": [
            "text"
          ],
          "cost": {
            "input": 0,
            "output": 0,
            "cacheRead": 0,
            "cacheWrite": 0
          },
          "api": "openai-completions"
        }
      ]
    }
  }
}

Check that the model is assigned to the agent in ~/.openclaw/openclaw.json. The workspace and agentDir is dependent on your setup.

    "list": [
      {
        "id": "main"
      },
      {
        "id": "sea-lion",
        "name": "sea-lion",
        "workspace": "/home/openclaw/.openclaw/workspace-sea-lion",
        "agentDir": "/home/openclaw/.openclaw/agents/sea-lion/agent",
        "model": "sealion/aisingapore/Qwen-SEA-LION-v4-32B-IT"
      }
    ]

These are several potential SEA-LION-powered features in your personal AI assistant:

  • SEA language translation
    • Users can send messages in English and receive the translations in Bahasa Indonesia, Thai, Vietnamese, Burmese, Malay, Filipino, or Tamil, and vice versa.
    • Useful for communicating across language barriers.
  • SEA travel
    • Helps plan trips across Southeast Asia with local-language recommendations for transport, food, customs, and safety (e.g. “3-day halal-friendly itinerary in Ho Chi Minh City, answer in Malay”).
    • Provides culturally-aware guidance on dress codes, festivals, taboos, and simple phrase translations so travellers can navigate confidently.
  • Cultural etiquette
    • Ask about local customs, holidays, or do/don’t etiquette (e.g. “What’s appropriate business etiquette in Bangkok?”).
    • Helpful for expats and remote teams working with SEA partners.

One of the reasons OpenClaw has grown so quickly is its flexibility: every user can configure a personal AI assistant tailored to their own workflows (for example, auto-responding to chats or emails, updating calendars, sending reminders, and more). You can use SEA-LION to power these customised agents, so your setup reflects your unique needs while remaining cost-effective.

Use SEA-LION as the Heartbeat model

OpenClaw’s Heartbeat is one of its most useful feature: it runs every 30 minutes (configurable) by default, scanning your workspace for tasks that need attention, kicking off periodic checks, and keeping your automations alive.

However, if a billable model is used to power the heartbeat, the frequent calls can add up to a lot of tokens and a costly bill.

Using a SEA-LION model delivers meaningful cost savings and keeps your automations sustainable even under frequent Heartbeat runs. Configure the heartbeat in agents.defaults (~/.openclaw/openclaw.json) to use a SEA-LION model, example:

{
  "heartbeat": {
    "every": "30m",
    "model": "sealion/aisingapore/Qwen-SEA-LION-v4-32B-IT",
    "includeReasoning": false,
    "target": "last",
    "prompt": "Read HEARTBEAT.md if it exists (workspace context). Follow it strictly. Do not infer or repeat old tasks from prior chats. If nothing needs attention, reply HEARTBEAT_OK.",
    "ackMaxChars": 300
  }
}

Conclusion

With SEA-LION wired into OpenClaw, you can:

  • Give users a multilingual, SEA-aware assistant that runs on their own devices
  • Let the heartbeat run often without worrying about potential bill shocks

OpenClaw evolves quickly, so the steps for adding custom models may change over time. Nonetheless, the SEA-LION models can be added using the model providers that support OpenAI-compatible APIs.

References