Documentation

Create your account

Go to subaiya.com/register and sign up with your email. You'll receive a verification link — click it to activate your account.

Copy your API key

After verification, you'll see your API key once. Copy it immediately and store it safely — it won't be shown again.

Configure OpenClaw

Add Subaiya as a provider in your OpenClaw config. Open the config file:

Add the following inside the "models" section. If no "models" key exists, add it at the top level:

"models": {
  "mode": "merge",
  "providers": {
    "custom-api-subaiya-com": {
      "baseUrl": "https://api.subaiya.com/t/YOUR-SUBAIYA-KEY",
      "api": "anthropic-messages",
      "apiKey": "YOUR-ANTHROPIC-KEY",
      "models": [
        {
          "id": "claude-sonnet-4-6",
          "name": "Claude Sonnet 4.6 via Subaiya",
          "contextWindow": 200000,
          "maxTokens": 8192,
          "input": ["text"]
        }
      ]
    }
  }
}

Replace YOUR-SUBAIYA-KEY with your sk-sub-... key from Step 2, and YOUR-ANTHROPIC-KEY with your Anthropic API key (sk-ant-...).

Using GPT models with a Codex subscription?

Add this as a second provider inside "providers":

"openai-codex": {
  "baseUrl": "https://api.subaiya.com/t/YOUR-SUBAIYA-KEY",
  "api": "openai-codex-responses",
  "models": [
    {
      "id": "gpt-5.4",
      "name": "GPT-5.4 via Subaiya",
      "contextWindow": 200000,
      "maxTokens": 16384,
      "input": ["text", "image"]
    }
  ]
}

⚠️ The "api": "openai-codex-responses" field is required since OpenClaw v2026.4.5. Without it, GPT requests may be routed incorrectly.

After saving, restart the gateway:

Set up your agents & models

Configure all models you want to use (e.g. GPT-5.4, Claude Sonnet, Haiku) in your OpenClaw config. Test each one to make sure it works through Subaiya. Customize your SOUL.md, IDENTITY.md, and other agent files as needed.

Take your time — the integrity monitor is inactive until you calibrate it.

Calibrate the integrity monitor

Once everything is set up, go to subaiya.com/dashboard and click "Calibrate" in the Integrity Monitor section.

Subaiya will learn your system's baseline — it monitors SOUL.md, AGENTS.md, and IDENTITY.md over the next 3 messages.

After 3 messages, the monitor is active and watches for unauthorized changes to your agent's identity files.

💡 Tip: If you add new models or change agent files later, click Calibrate again to set a new baseline.

Prompt Injection Detection

Subaiya scans for hidden instructions in files, emails, websites, and messages that try to manipulate your agent. When an attempt is detected, you'll see an alert in your dashboard showing the source, type of attack, and details.

File Integrity Monitoring

Critical files like SOUL.md, AGENTS.md, and IDENTITY.md are monitored for unauthorized changes. If a file is modified, deleted, or a new file appears, you'll get an alert. Click "Confirm changes" to accept intentional updates and set a new baseline.

Media Warnings

When your agent receives images or media files, Subaiya alerts you in the dashboard. Images can potentially contain hidden instructions. Your agent is instructed to treat all media as information only — not as commands.

Group Chat Protection

When using your agent in group chats (WhatsApp, Telegram, Slack, Discord), Subaiya can detect message senders and instruct your agent to only follow commands from you. Requires configuration. Contact us for setup help.

Budget Control

Set a spending limit per session ($5, $10, $20, $50, or custom). Track API costs in real-time with USD/EUR toggle. When the limit is reached, the agent pauses and asks you to raise it or start a new session. Only applies to pay-per-token usage.

Emergency Stop

The red STOP AGENT button at the top of your dashboard instantly blocks all agent actions. Press it again to resume. After stopping, start a new conversation in OpenClaw for a clean restart.