← Back to Docs
🦞

Connect OpenClaw

Save 40-60% on every LLM call through OpenClaw

OpenClaw is a self-hosted gateway that connects chat apps (WhatsApp, Telegram, Discord) to AI coding agents. AgentReady works as a custom model provider — every message your agent processes gets compressed automatically, saving 40-60% on tokens.

How It Works

💬

Chat normally

Message your agent on WhatsApp, Telegram, Discord — nothing changes

✂️

Auto-compressed

AgentReady compresses every prompt before it hits the LLM. Same meaning, fewer tokens

💰

Save 40-60%

Your Anthropic/OpenAI bill drops by half. ~5ms overhead, 0.4% accuracy delta

Prerequisites

  1. OpenClaw installed and running (openclaw gateway)
  2. An AgentReady API key — get one free in 30 seconds
  3. An OpenAI API key (you probably already have this)
1

Add AgentReady as a Custom Provider

Add this to your ~/.openclaw/openclaw.json. OpenClaw's hot-reload will pick it up — no restart needed.

~/.openclaw/openclaw.json (merge into existing config)
{
  models: {
    providers: {
      agentready: {
        baseUrl: "https://agentready.cloud/v1",
        apiKey: "ak_your_agentready_key",
        api: "openai-completions",
        headers: { "X-Upstream-API-Key": "${OPENAI_API_KEY}" },
        models: [
          {
            id: "gpt-4o",
            name: "GPT-4o via AgentReady (−60% tokens)",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 128000,
            maxTokens: 16384,
          },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: { primary: "agentready/gpt-4o" },
    },
  },
}
2

That's It — Test It

Send a message to your agent through any channel. Check the OpenClaw Control UI — you'll see requests going through agentready/gpt-4o. Your prompts are now automatically compressed.

💡 Tip: Use /model agentready/gpt-4o in any chat to switch to AgentReady on-the-fly, or set it as default in the config above.

⚙️

Advanced: Multiple Models + Aliases

Route multiple OpenAI models through AgentReady with short aliases:

Multi-model config
{
  models: {
    providers: {
      agentready: {
        baseUrl: "https://agentready.cloud/v1",
        apiKey: "ak_your_agentready_key",
        api: "openai-completions",
        headers: { "X-Upstream-API-Key": "${OPENAI_API_KEY}" },
        models: [
          {
            id: "gpt-4o",
            name: "GPT-4o via AgentReady",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 128000,
            maxTokens: 16384,
          },
          {
            id: "gpt-4o-mini",
            name: "GPT-4o Mini via AgentReady",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 128000,
            maxTokens: 16384,
          },
          {
            id: "gpt-4.1",
            name: "GPT-4.1 via AgentReady",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 128000,
            maxTokens: 32768,
          },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: {
        primary: "agentready/gpt-4o",
        fallbacks: ["agentready/gpt-4o-mini"],
      },
      models: {
        "agentready/gpt-4o": { alias: "ar" },
        "agentready/gpt-4o-mini": { alias: "ar-mini" },
      },
    },
  },
}

Now you can type /model ar or /model ar-mini in chat to switch.

🔐

Keep Keys Safe with Env Vars

Don't hardcode keys. Use ~/.openclaw/.env + variable substitution:

~/.openclaw/.env
# Add to ~/.openclaw/.env
AGENTREADY_API_KEY=ak_your_agentready_key
OPENAI_API_KEY=sk-your_openai_key
~/.openclaw/openclaw.json (with env vars)
{
  env: {
    AGENTREADY_API_KEY: "ak_your_agentready_key",
  },
  models: {
    providers: {
      agentready: {
        baseUrl: "https://agentready.cloud/v1",
        apiKey: "${AGENTREADY_API_KEY}",
        api: "openai-completions",
        headers: { "X-Upstream-API-Key": "${OPENAI_API_KEY}" },
        models: [
          {
            id: "gpt-4o",
            name: "GPT-4o via AgentReady (−60% tokens)",
            reasoning: false,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 128000,
            maxTokens: 16384,
          },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: { primary: "agentready/gpt-4o" },
    },
  },
}

40-60% saved

~5ms overhead

0.4% accuracy Δ

Hot reload

Ready to start?

Get your free API key and paste the config — takes 30 seconds.

Get Free API Key OpenClaw Docs

FAQ

Does it work with Anthropic models too?

AgentReady currently proxies OpenAI-compatible endpoints. For Anthropic models, the compression works on the prompt text — just route your OpenAI calls through AgentReady while keeping Anthropic direct.

Will it slow down my agent?

No. AgentReady adds ~5ms per request. On a typical GPT-4o call that takes 2-10 seconds, this is imperceptible.

Is it free?

Yes — AgentReady is completely free during the open beta. No credit card required, no usage limits.

Do I need to restart OpenClaw?

No. OpenClaw hot-reloads model and provider config changes. Just save the file and it picks up the new provider automatically.