A decision lab your business owns
People and agents work the same Problems. The council returns a cited recommendation with its strongest dissent, a named owner takes the decision, and on the review date the outcome is checked against the prediction. People and agents work the same Problems. Every decision gets a named owner, a review date and a checked outcome.
Applied
- Council recommendation
- Austria first, with a staged rollout
- Strongest dissent
- Channel data is thin; test before committing budget
- Decision owner
- Head of Growth · person
- Proposed by
- Claude Code · agent, applied by a person
- Prediction
- 20 paying accounts within 90 days · confidence 0.6
- Review date
- 15 Jan 2027
From question to checked outcome
A Problem keeps its question, evidence, council runs, decisions and outcomes together for its whole life, so the reasoning is still there when you review the result.
Open a Problem
Write the question, the options and the stakes, and attach evidence. People and agents can both add to it.
Convene the council
Six AI advisory perspectives compare the options and return a cited recommendation with the strongest dissent. Every advisor's reasoning stays readable.
Record the decision
Name the owner, person or agent, with the rationale, evidence, a prediction and a review date. It stays proposed until a person applies it.
Check the outcome
On the review date, record what happened and mark the prediction as a hit, partial or miss. The whole responsibility trail stays in one view and one API call.
People and agents operate the same lab
The web app, the hosted MCP server and in-page WebMCP tools all work on the same tenant-scoped state. Nothing an agent does is hidden from the people who govern it.
People: the web app
Open a Problem, inspect evidence and contradictions, read each advisor's reasoning, and apply, decline or withdraw every proposed action.
Agents: hosted MCP
Claude Code, Codex, Cursor, Gemini CLI, OpenCode or your own MCP client can open Problems, convene the council, record decisions and propose outcomes. Agent actions stay proposed until a person applies them.
read -s COUNCILNOW_API_KEY && export COUNCILNOW_API_KEY
claude mcp add --transport http councilnow https://councilnow.com/mcp --header "Authorization: Bearer ${COUNCILNOW_API_KEY}"Create an API key in the app under Settings → Connect agent. The first line reads the key without echoing it or saving it to shell history; the second connects Claude Code.
Your decisions stay yours
No required copilot, no single model vendor, and no proprietary format between your business and its own judgment.
Open models first
We prefer open-weight models and use closed models where a task needs them.
Privacy tiers
Standard uses any approved model. Private limits a workspace or a single Problem to open-weight models on zero-retention, no-training providers in the regions you accept. A tier can be made stricter, never silently weaker.
Any MCP agent
Claude Code, Codex, Cursor, Gemini CLI, OpenCode or your own client. No copilot is required, and agents use the same entitled API as the web app.
Take your data with you
Export every Problem with its full event history as JSON through the API.
Monthly plans from €9
Higher plans include more usage. Plans are listed at shop.verduona.com, which is a preview: online ordering is not open yet. Access is by invitation or review, and payment alone does not activate a workspace.
For a product or growth team making repeated calls every week.
View plan in shopFor leadership teams running linked decisions across the business.
View plan in shopFor organisations that need agreed rollout, security review and procurement support.
Contact Sales or view the plan in the shopIncluded in every plan
- Web app and hosted MCP for your agents
- Problems with evidence, decisions, owners and outcomes
- Standard and private privacy tiers
- JSON export of your Problems
Six AI perspectives on every decision
The six advisors are AI advisory perspectives, not human professional advisers. They compare the same options and prepare decision-support material; a named person or agent owns the decision.
The Strategist
Tests which option creates the strongest position 12 to 24 months out and where the recommendation earns leverage.
The Pragmatist
Flags the delivery bottleneck, resource drag, and what must already be true for the lead option to work.
The Contrarian
Builds the strongest case against the lead option so the final call includes a real reversal trigger, not fake consensus.
The Customer Voice
Checks which option customers will actually feel, what behavior must change, and where the decision could lose trust.
The Operator
Turns the lead option into the next move, naming dependencies, owners, and what has to happen first.
The Analyst
Quantifies upside, downside, and confidence so the recommendation is numerically defensible, not just persuasive.
Run your next decision through a lab you own
Start with one Problem, then bring your team and your agents. Check the plan and access conditions before ordering; payment does not activate a workspace.
Monthly plans start at €9. Access is confirmed separately from payment.