Guide to Using OpenClaw with 开桥 ​

OpenClaw is a personal AI assistant that runs on your own machine and answers you on the channels you already use (terminal TUI, web dashboard, WhatsApp, Telegram, Discord, Slack, Signal, iMessage, and more). With the published @openqiao/openclaw-openqiao-provider plugin on ClawHub, a single install command gives OpenClaw access to all 135+ 开桥 models — GPT, Claude, Gemini, Grok, DeepSeek, Qwen, and more — with on-demand dynamic resolution and a one-time per-host catalog cache.

Prerequisites ​

• Node.js 22 or later
• OpenClaw >= 2026.4.14 (openclaw --version)
• A 开桥 API key — see Step 0 below

Step 0: Get a 开桥 API Key ​

开桥 offers two billing options. Either works with this plugin.

Best for: personal development, learning/exploration
Features: fixed monthly fee, predictable cost, 5-10x price leverage
API Key format: sk-ss-v1-xxx

How to get it:
1. Visit the subscription page: https://openqiao.com/platform/subscription
2. Choose a plan (Starter $20/month, Max $100/month, Ultra $200/month)
3. After subscribing, create a subscription API Key on the page

For details, see: Subscription Plan Guide
https://docs.openqiao.com/guide/subscription

Best for: production environments, commercial products, enterprise apps
Features: no rate limits, production-grade stability, billed by actual usage
API Key format: sk-ai-v1-xxx

How to get it:
1. Visit the pay-as-you-go page: https://openqiao.com/platform/pay-as-you-go
2. Top up your account
3. Create an API Key in the "Pay As You Go API Keys" section

For details, see: Pay-as-you-go Guide
https://docs.openqiao.com/guide/pay-as-you-go

Quick Start (3 commands) ​

If you already have OpenClaw and your API key set, this is the entire setup:

export OPENQIAO_API_KEY=sk-ss-v1-...

openclaw plugins install clawhub:@openqiao/openclaw-openqiao-provider
openclaw onboard --openqiao-api-key "$OPENQIAO_API_KEY" --auth-choice openqiao-api-key \
  --flow manual --accept-risk --non-interactive
openclaw gateway install && openclaw gateway start

Then openclaw chat to start chatting. The default model is openqiao/openai/gpt-5.4; switch to any other 开桥 model in chat with /model openqiao/<id>.

The detailed walkthrough below explains each step and how to verify it.

Method 1: Install via ClawHub plugin (Recommended) ​

Step 1: Verify or upgrade OpenClaw ​

openclaw --version            # expect: OpenClaw 2026.4.14 or newer

If older, upgrade:

npm i -g openclaw@latest
openclaw --version            # confirm >= 2026.4.14

If npm i -g errors with permissions, you're on system Node — switch to a Node version manager (nvm/volta/fnm) or prefix with sudo.

Step 2: Set your 开桥 API key ​

export OPENQIAO_API_KEY=sk-ss-v1-...

Step 3: Install the plugin from ClawHub ​

openclaw plugins install clawhub:@openqiao/openclaw-openqiao-provider

You'll see output similar to:

ClawHub code-plugin @openqiao/openclaw-openqiao-provider@0.2.1 channel=community verification=source-linked
ClawHub package "@openqiao/openclaw-openqiao-provider" is community; review source and verification before enabling.
Downloading plugin @openqiao/openclaw-openqiao-provider@0.2.1 from ClawHub…
Installing to ~/.openclaw/extensions/openqiao…
Installed plugin: openqiao
Restart the gateway to load plugins.

The "community plugin" notice is a standard advisory for any non-bundled plugin and is expected.

Step 4: Onboard with your API key (do this before starting the gateway) ​

This step writes the gateway mode, the provider config, and the auth profile. The OpenClaw gateway daemon refuses to start without gateway.mode set, so onboard must run first.

openclaw onboard \
  --openqiao-api-key "$OPENQIAO_API_KEY" \
  --auth-choice openqiao-api-key \
  --flow manual \
  --accept-risk \
  --non-interactive

Expected output:

Updated ~/.openclaw/openclaw.json
Workspace OK: ~/.openclaw/workspace
Sessions OK: ~/.openclaw/agents/main/sessions
Gateway did not become reachable at ws://127.0.0.1:18789.

The Gateway did not become reachable line is informational — onboard ran a post-write health check; the gateway will be installed in the next step.

Step 5: Install + start the gateway service ​

openclaw gateway install
openclaw gateway start
sleep 5
openclaw gateway status | grep -E "Runtime:|Listening|Connectivity probe|Capability"

Expected:

Runtime: running (pid …)
Listening: 127.0.0.1:18789
Connectivity probe: ok
Capability: admin-capable

Step 6: Verify the configuration ​

node -e "
const c = require(require('os').homedir() + '/.openclaw/openclaw.json');
const p = require(require('os').homedir() + '/.openclaw/extensions/openqiao/package.json');
console.log('plugin name    :', p.name);
console.log('plugin version :', p.version);
console.log('primary model  :', c.agents?.defaults?.model?.primary);
console.log('allowlist      :', JSON.stringify(c.agents?.defaults?.models));
console.log('openqiao api     :', c.models?.providers?.openqiao?.api);
console.log('openqiao baseUrl :', c.models?.providers?.openqiao?.baseUrl);
console.log('plugin enabled :', c.plugins?.entries?.openqiao?.enabled);
console.log('gateway.mode   :', c.gateway?.mode);
console.log('auth profile   :', c.auth?.profiles?.['openqiao:default']?.mode);
"

Expected output:

plugin name    : @openqiao/openclaw-openqiao-provider
plugin version : 0.2.1
primary model  : openqiao/openai/gpt-5.4
allowlist      : {}
openqiao api     : openai-completions
openqiao baseUrl : https://openqiao.com/api/v1
plugin enabled : true
gateway.mode   : local
auth profile   : api_key

The empty allowlist: {} is the key signal — any openqiao/<id> is selectable, no manual pinning required.

Step 7: Smoke test through three vendors ​

# Default model — fast, no cache fetch needed
openclaw infer model run --gateway --model openqiao/openai/gpt-5.4 \
  --prompt "Reply with: DEFAULT_OK"

# Anthropic via 开桥 — first call triggers a one-time ~1-2s catalog fetch
openclaw infer model run --gateway --model openqiao/anthropic/claude-opus-4.7 \
  --prompt "Reply with: FREE_PICK"

# xAI via 开桥 — cache is now warm, instant
openclaw infer model run --gateway --model openqiao/x-ai/grok-4.3 \
  --prompt "Reply with: GROK_OK"

Each should print provider: openqiao, the model id, and a short reply. No model not allowed errors.

Step 8: Open chat ​

openclaw chat                 # local terminal UI
# or
openclaw dashboard            # web UI in browser

You're done. Day-to-day usage is below.

Using 开桥 models in chat ​

Switching models on the fly ​

In both the terminal TUI (openclaw chat) and the web dashboard chat input, type a slash directive:

/model openqiao/anthropic/claude-opus-4.7    switch to Claude Opus 4.7
/model openqiao/x-ai/grok-4.3                switch to Grok 4.3
/model openqiao/deepseek/deepseek-v4-pro     switch to DeepSeek v4 Pro
/model openqiao/qwen/qwen3.6-max-preview     switch to Qwen 3.6 Max
/model status                              show current model + endpoint

Any of the 135+ 开桥 models works directly — no need to pre-pin. The first call to a previously-unused model triggers a one-time ~1-2s catalog warmup; subsequent calls hit the disk cache at ~/.openclaw/cache/openqiao-models.json and are instant.

Listing available models ​

# All 开桥 models from the live catalog
curl -s https://openqiao.com/api/v1/models | python3 -c "
import sys, json
for m in json.load(sys.stdin)['data']: print(m['id'])
"

# Or browse via the official model list page
open https://openqiao.com/models

One-shot CLI inference ​

openclaw infer model run --gateway --model openqiao/anthropic/claude-opus-4.7 \
  --prompt "Summarize the README in 3 bullets"

Pinning favorites in the dashboard dropdown (optional) ​

The dashboard chat-header dropdown shows only your saved aliases. To populate it with a few favorites:

openclaw models aliases add Opus  openqiao/anthropic/claude-opus-4.7
openclaw models aliases add GPT55 openqiao/openai/gpt-5.5-pro
openclaw models aliases add Grok  openqiao/x-ai/grok-4.3
openclaw gateway restart

Method 2: Manual configuration (fallback) ​

Step 1: Install OpenClaw ​

npm install -g openclaw@latest
openclaw setup --mode local --non-interactive   # writes baseline config

Step 2: Edit ~/.openclaw/openclaw.json ​

Add the 开桥 provider with the models you want:

{
  "models": {
    "mode": "merge",
    "providers": {
      "openqiao": {
        "baseUrl": "https://openqiao.com/api/v1",
        "apiKey": "sk-ss-v1-your-api-key-here",
        "api": "openai-completions",
        "models": [
          {
            "id": "openai/gpt-5.4",
            "name": "GPT-5.4 via 开桥",
            "reasoning": false,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 200000,
            "maxTokens": 8192
          },
          {
            "id": "anthropic/claude-opus-4.7",
            "name": "Claude Opus 4.7 via 开桥",
            "reasoning": true,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 200000,
            "maxTokens": 8192
          },
          {
            "id": "google/gemini-3.1-pro-preview",
            "name": "Gemini 3.1 Pro via 开桥",
            "reasoning": false,
            "input": ["text", "image"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 1048576,
            "maxTokens": 8192
          },
          {
            "id": "deepseek/deepseek-v4-pro",
            "name": "DeepSeek v4 Pro via 开桥",
            "reasoning": false,
            "input": ["text"],
            "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
            "contextWindow": 128000,
            "maxTokens": 8192
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "openqiao/openai/gpt-5.4"
      }
    }
  }
}

Step 3: Start the gateway ​

openclaw gateway install
openclaw gateway restart

Step 4: Verify ​

openclaw infer model providers --json | python3 -c "
import sys, json
ps = json.load(sys.stdin)
print(next((p for p in ps if p.get('provider')=='openqiao'), None))
"

openclaw infer model run --gateway --model openqiao/openai/gpt-5.4 \
  --prompt "Reply: OK"

Adding more models ​

Add entries to the models array. Field reference:

After editing, restart the gateway.

Troubleshooting ​

openclaw onboard --openqiao-api-key "$OPENQIAO_API_KEY" --auth-choice openqiao-api-key \
  --flow manual --accept-risk --non-interactive
openclaw gateway start

openclaw plugins update openqiao       # pulls 0.2.0+
openclaw plugins inspect openqiao | grep Version

node -e "
const fs=require('fs');
const path=require('os').homedir()+'/.openclaw/openclaw.json';
const c=JSON.parse(fs.readFileSync(path,'utf8'));
if (c.agents?.defaults) c.agents.defaults.models = {};
fs.writeFileSync(path, JSON.stringify(c, null, 2));
"
openclaw gateway restart

openclaw plugins inspect openqiao
# Status should be "loaded"
# Source should be "clawhub"
# Recorded version should be 0.2.1 (or newer)

openclaw plugins uninstall openqiao
openclaw plugins install clawhub:@openqiao/openclaw-openqiao-provider
openclaw gateway restart

Supported Models ​

开桥 provides access to 135+ models. Some popular options:

For the live full list, visit openqiao.com/models. New models that 开桥 adds are usable the moment they ship — /model openqiao/<new-id> works without a plugin update.

Contact Us ​