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:
QUORUMAI_API_KEY=qai_your_key_here
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)
curl -s -X POST https://quorumai.io/api/v1/inquiry \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $QUORUMAI_API_KEY" \
-d '{
"thesis": "USER_QUESTION_HERE",
"mode": "quick",
"academy": "general"
}'
Standard Inquiry (dedicated arbiter, better scoring)
curl -s -X POST https://quorumai.io/api/v1/inquiry \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $QUORUMAI_API_KEY" \
-d '{
"thesis": "USER_QUESTION_HERE",
"mode": "standard",
"academy": "general"
}'
Deep Inquiry (Pro only — 4 rounds, 16 responses)
curl -s -X POST https://quorumai.io/api/v1/inquiry \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $QUORUMAI_API_KEY" \
-d '{
"thesis": "USER_QUESTION_HERE",
"mode": "deep",
"academy": "general",
"dissent": true
}'
Parameters
| Field | Type | Default | Description |
|---|---|---|---|
thesis | string | required | The question to ask (alias: question) |
academy | 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 |
|---|---|
synthesis | The combined best answer — always show this first |
prima | 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
- 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:
🏛️ QuorumAI Synthesis:
[synthesis text]
🏆 Top Voice: [prima provider] ([prima score])
🔗 Full result: [share_url]
If dissent mode was used, add:
✅ Consensus: [consensus_claims]
⚠️ Vulnerable: [vulnerable_claims]
📊 Confidence: [confidence]%
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
USER_QUESTION_HERE - Default to
quickmode andgeneralacademy unless the user specifies otherwise - If the user says "deep dive" or "thorough analysis", use
deepmode - If the user mentions medical/health topics, suggest
clinicalacademy - If the user mentions business/strategy, suggest
businessacademy - If the user mentions code/engineering, suggest
technicalacademy - If the user mentions investing/finance, suggest
investmentacademy - 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