QuorumAI — Multi-Model AI Synthesis
Ask a question once. Four AI models (Claude, GPT, Gemini, Grok) answer simultaneously from specialized perspectives, an arbiter scores them, and you get one synthesized best answer.
Setup
Get your free API key at https://quorumai.io/account (sign up, go to API Keys, generate one). The key starts with qai_.
Set it in your OpenClaw config:
CODEBLOCK0
Usage
When the user asks you to use QuorumAI, run a quorumai inquiry, or says "ask quorumai", send their question to the QuorumAI API.
Quick Inquiry (default)
CODEBLOCK1
Standard Inquiry (dedicated arbiter, better scoring)
CODEBLOCK2
Deep Inquiry (Pro only — 4 rounds, 16 responses)
CODEBLOCK3
Parameters
| Field | Type | Default | Description |
|---|
| INLINECODE1 | string | required | The question to ask (alias: question) |
| INLINECODE3 |
string |
"general" | Academy to use:
general,
clinical,
business,
technical,
investment |
|
mode | string |
"quick" | Inquiry depth:
quick (Free),
standard (Free),
deep (Pro) |
|
arbiter | string |
"auto" | Which model judges:
auto,
claude,
gpt,
gemini,
grok |
|
dissent | boolean |
false | Enable dissent mode — models argue against themselves to stress-test claims (Pro) |
Academies
- - general — General-purpose inquiry with balanced perspectives
- clinical — Clinical Review Panel for medical and health questions
- business — Business Advisory Board for strategy and operations
- technical — Technical Review Board for engineering and architecture
- investment — Investment Committee for financial analysis
Response Format
The API returns JSON. Here are the key fields to present to the user:
| Field | Description |
|---|
| INLINECODE24 | The combined best answer — always show this first |
| INLINECODE25 |
The top-scoring voice (winning model) |
|
voices | Array of individual model responses with provider and school (role) |
|
scores | Arbiter's scoring of each voice |
|
confidence | Overall confidence percentage (when dissent is enabled) |
|
consensus_claims | Claims all models agreed on (dissent mode) |
|
vulnerable_claims | Claims that didn't survive adversarial analysis (dissent mode) |
|
share_url | Public link to view the full inquiry result |
How to Present Results
- 1. Always lead with the synthesis — this is the main answer
- Show the prima (winning voice) and its score
- If the user wants detail, show individual voice responses
- If dissent was enabled, highlight consensus vs vulnerable claims
- Always include the share_url so the user can view the full result on quorumai.io
Example output formatting:
CODEBLOCK4
If dissent mode was used, add:
CODEBLOCK5
Rate Limits
- - Free tier: 5 inquiries/day (Quick and Standard modes only)
- Pro tier ($9.99/mo): 50 inquiries/day, all modes, dissent, all academies
Rate limit headers are returned: X-RateLimit-Limit and X-RateLimit-Remaining.
If the user hits a rate limit, suggest upgrading at https://quorumai.io/account
Error Handling
- - 401: Invalid or missing API key — direct user to https://quorumai.io/account
- 403: Feature requires Pro tier — suggest upgrade
- 429: Rate limit exceeded — show remaining reset time
- 500: Server error — retry once, then report the issue
Rules
- - Always substitute the user's actual question for INLINECODE34
- Default to
quick mode and general academy unless the user specifies otherwise - If the user says "deep dive" or "thorough analysis", use
deep mode - If the user mentions medical/health topics, suggest
clinical academy - If the user mentions business/strategy, suggest
business academy - If the user mentions code/engineering, suggest
technical academy - If the user mentions investing/finance, suggest
investment academy - Never expose the raw API key in messages to the user
- Responses can take 15-45 seconds depending on mode — let the user know it's processing
QuorumAI — 多模型AI综合系统
只需提出一个问题。四个AI模型（Claude、GPT、Gemini、Grok）将从专业角度同时作答，由仲裁者评分，最终生成一个综合最优答案。
配置
在 https://quorumai.io/account 获取免费API密钥（注册后进入API Keys页面生成）。密钥以 qai_ 开头。
在OpenClaw配置中设置：
QUORUMAIAPIKEY=qaiyourkey_here
使用方法
当用户要求你使用QuorumAI、发起quorumai查询或说ask quorumai时，将用户的问题发送至QuorumAI API。
快速查询（默认）
bash
curl -s -X POST https://quorumai.io/api/v1/inquiry \
-H Content-Type: application/json \
-H Authorization: Bearer $QUORUMAIAPIKEY \
-d {
thesis: 用户问题,
mode: quick,
academy: general
}
标准查询（专用仲裁者，更优评分）
bash
curl -s -X POST https://quorumai.io/api/v1/inquiry \
-H Content-Type: application/json \
-H Authorization: Bearer $QUORUMAIAPIKEY \
-d {
thesis: 用户问题,
mode: standard,
academy: general
}
深度查询（仅Pro版 — 4轮，16次回答）
bash
curl -s -X POST https://quorumai.io/api/v1/inquiry \
-H Content-Type: application/json \
-H Authorization: Bearer $QUORUMAIAPIKEY \
-d {
thesis: 用户问题,
mode: deep,
academy: general,
dissent: true
}
参数
| 字段 | 类型 | 默认值 | 描述 |
|---|
| thesis | 字符串 | 必填 | 要提出的问题（别名：question） |
| academy |
字符串 | general | 使用的学院：general、clinical、business、technical、investment |
| mode | 字符串 | quick | 查询深度：quick（免费）、standard（免费）、deep（Pro） |
| arbiter | 字符串 | auto | 评判模型：auto、claude、gpt、gemini、grok |
| dissent | 布尔值 | false | 启用异议模式 — 模型自我反驳以压力测试观点（Pro版） |
学院
- - general — 通用查询，视角均衡
- clinical — 临床评审委员会，适用于医疗健康问题
- business — 商业顾问委员会，适用于战略与运营
- technical — 技术评审委员会，适用于工程与架构
- investment — 投资委员会，适用于财务分析
响应格式
API返回JSON格式。以下是需要呈现给用户的关键字段：
| 字段 | 描述 |
|---|
| synthesis | 综合最优答案 — 始终优先展示此项 |
| prima |
得分最高的声音（获胜模型） |
| voices | 各模型回答数组，包含提供商和学派（角色） |
| scores | 仲裁者对每个声音的评分 |
| confidence | 总体置信度百分比（启用异议模式时） |
| consensus_claims | 所有模型达成一致的论点（异议模式） |
| vulnerable_claims | 未通过对抗性分析的论点（异议模式） |
| share_url | 查看完整查询结果的公开链接 |
结果呈现方式
- 1. 始终以综合结果开头 — 这是主要答案
- 展示最优声音（获胜模型）及其评分
- 如果用户需要详情，展示各模型的回答
- 如果启用了异议模式，突出展示共识论点与脆弱论点
- 始终包含share_url，方便用户在quorumai.io上查看完整结果
输出格式示例：
🏛️ QuorumAI综合结果：
[综合文本]
🏆 最优声音：[最优模型提供商]（[最优评分]）
🔗 完整结果：[分享链接]
如果使用了异议模式，添加：
✅ 共识论点：[共识论点]
⚠️ 脆弱论点：[脆弱论点]
📊 置信度：[置信度]%
速率限制
- - 免费版：每天5次查询（仅限快速和标准模式）
- Pro版（$9.99/月）：每天50次查询，支持所有模式、异议模式、所有学院
速率限制头部信息会返回：X-RateLimit-Limit 和 X-RateLimit-Remaining。
如果用户达到速率限制，建议升级至 https://quorumai.io/account
错误处理
- - 401：API密钥无效或缺失 — 引导用户访问 https://quorumai.io/account
- 403：功能需要Pro版 — 建议升级
- 429：超出速率限制 — 显示剩余重置时间
- 500：服务器错误 — 重试一次，然后报告问题
规则
- - 始终将用户的真实问题替换 用户问题
- 除非用户另有指定，默认使用 quick 模式和 general 学院
- 如果用户说深度探索或全面分析，使用 deep 模式
- 如果用户提及医疗/健康话题，建议使用 clinical 学院
- 如果用户提及商业/战略，建议使用 business 学院
- 如果用户提及代码/工程，建议使用 technical 学院
- 如果用户提及投资/金融，建议使用 investment 学院
- 切勿在给用户的消息中暴露原始API密钥
- 根据模式不同，响应可能需要15-45秒 — 告知用户正在处理中
