Built on the leading frontier models
Five AIs debate. One Decision Brief.
Question
We're building a fintech tool. Should we focus on the underwriting workflow or the marketing layer first?
Final consensus
Waiting for the council to finish…
GPT, Claude, Gemini, Grok, and Perplexity each weigh in. They debate. They name where they disagree. Then you get one structured Decision Brief: verdict, confidence, dissent, next step. The kind of document you can actually ship to your team.
The before / after
One AI gives you an answer.
The council gives you a decision.
A single model will always sound confident. The council shows you the cracks: the dissent, the unstated assumptions, the risks one mind would have buried in a fluent paragraph.
Output
"Great question! There are several factors to consider here. You could either raise prices on your existing customers or expand to new geographic markets first. Both approaches have merit. Ultimately, the right answer depends on your specific situation. I hope this helps!"
- Confident tone, low-signal answer
- No risks, no dissent, no math
- You can't share this to a team
Verdict
Raise on net-new accounts, hold pricing on existing for 1 quarter, then re-price after retention data.
Risks
Churn lag will hide damage for 60 days
Dissent
Sonar Pro: hold flat, expand first
Assumption
NRR >110% on existing book
Next step
Run cohort survey before pricing change
- Quantified, structured, and act-on-able
- Disagreements surfaced, not hidden
- One click export to a decision brief
How it works
The answer has to survive the process.
Three phases. Independent, then adversarial, then synthesized. Each step is visible while it happens.
Independent answers
Every selected model answers in isolation. No blending, no first-mover bias. You see five distinct takes side by side.
Open debate
Models read each other's reasoning. They challenge gaps, weak math, vague claims, and missing context. Disagreement is encouraged, not hidden.
Decision brief
Claude Opus 4.7 synthesizes a structured brief: verdict, confidence, agreements, dissent, risks, assumptions, next steps. One artifact you can ship.
The artifact
Not a chat log. A Decision Brief.
Every run produces one structured brief: the verdict, a confidence score, the points of agreement, the dissent, the risks, the assumptions, the missing context, and the next step. Export to PDF, Markdown, or a shareable link.
Designed so a teammate can read the brief in 90 seconds and understand the call without rerunning anything.
Council Decision Brief
May 16 2026 · 5 models, 1 question
Question
Should we raise prices on existing customers, or focus capital on expanding to two new markets?
Verdict
Raise on net-new in Q3. Hold pricing on the existing book for a quarter, then re-price once NRR drag is measurable.
Agree
Net-new pricing first
Dissent
Sonar Pro: expand first
Risk
60d churn lag hides it
Next step
Cohort survey by Fri
Export to a sharable Decision Brief
Disagreement map
Where your 5 models split.
| Claim | |||||
|---|---|---|---|---|---|
| Build the underwriting workflow before marketing | |||||
| Demand test before infra spend | |||||
| Regulatory drag is the largest near term risk | |||||
| A document native interface is the wedge | |||||
| Pricing in the $499 to $999 ARPU lane |
Sell the disagreement, not the consensus. The split is where the real decision lives.
Sell the disagreement
Most AI tools hide the conflict.
We map it.
The Disagreement Map shows, per claim, exactly where your 5 models split. Strong agreement is comfortable. Mixed signal is the part you read twice. A clean split is where the real decision lives.
Models
13+
Roles
22+
Presets
13
What founders use it for
Pick a scenario. Watch the council run.
Six pre-wired councils for the decisions you actually make. One click, all roles assigned, all models picked, ready to go.
Stress test a decision
Devils advocate, customer voice, and skeptical investor pull at the same plan from three sides.
Pressure test a launch
Growth, brand critic, product manager, customer voice. See where the launch story actually lands.
Review architecture for security
Security auditor, performance optimizer, and code reviewer take the same diff from three angles.
Quantify a hire vs hire
Financial analyst, CFO advisor, COO advisor. One senior or two mid level. Math, not vibes.
Validate a business idea
Market researcher, customer voice, brand critic, financial analyst, skeptical investor. All in 45 seconds.
Audit a contract for risk
Legal risk reviewer, financial analyst, CFO advisor. Highlight the clauses that bite later.
Attach source material to ground the council
Document context
Drop in the deck. The council reads it before debating.
Upload PDFs, Word docs, spreadsheets, CSVs, Markdown, JSON, text, and code. The app extracts the readable content first, then makes every model use it as ground truth during the debate.
- Up to 8 files per run
- Up to 40 MB total
- Code, sheets, docs, PDFs
- Models cite the source
Founders use the council for
Why AI Council
Every other tool stops at the answer. We start at the decision.
Decision Brief, not chat
Verdict, confidence, agreements, dissent, risks, assumptions, next step. One artifact a teammate can read in 90 seconds.
Disagreement Map
We expose where the models split, per claim, with quotes. Nobody else has this. It's where the real risk lives.
Business Council presets
13 pre-wired advisory boards. Model picked, role assigned, prompt tuned. Click Run.
Document grounded
Upload the deck or contract. Every model uses it as ground truth before responding. No invented facts.
Real role personas
Skeptical Investor, Customer Voice, Security Auditor, Performance Optimizer. 22+ roles, each with a real prompt.
Export everywhere
PDF, Markdown, shareable link. Built so the brief travels: Slack, Notion, Linear, an exec email.
Pricing
Each run uses 5+ frontier models. Price reflects that.
Independent answers, real debate, synthesis. Plans keep the experience premium without forcing you into enterprise pricing.
Free
Try the council on a real question.
- 3 council runs / month
- Up to 5 models
- All 3 council modes
- PDF + doc upload
Council
For occasional strategic decisions.
- 20 council runs / month
- 15 image generations / month
- Chat with any model, together
- Full Decision Brief export
- Email support
Council Pro
For founders, operators, and builders running weekly councils.
- 60 council runs / month
- 30 image generations / month
- Chat with any model, together
- Per-answer export (PDF, Markdown)
- Priority synthesis model
- Priority support
Council Max
For teams running councils across departments.
- 200 council runs / month
- 250 image generations / month
- Chat with any model, together (4K image gen)
- Team seats + API access
- Dedicated support
One-off council runs are available for $1.99. No subscription required.
See all features and FAQsRun the Council from inside Claude.
Drop the AI Council into Claude Desktop or Claude Code as MCP tools. Call council, chat, and image_gen from any agent that speaks MCP. Same models, same persistence, no juggling provider keys.
council
5-model debate → brief
chat
Up to 5 models per turn
image_gen
OpenAI · Imagen · Grok
3 steps, ~2 minutes. Free with any AI Council account.
Ready to run your first council?
One free run. No credit card. 90 seconds.
Bring a real decision you're wrestling with. We'll bring the council.