wise-council-mcp
<div align="center">
# Wise Council MCP
**Bring multiple contrasting decision frameworks into one MCP-powered council.**
[English](./README.md) | [简体中文](./README.zh-CN.md)
[](https://modelcontextprotocol.io)
[](https://www.typescriptlang.org/)
[](./package.json)
[](#license)
</div>
---
## 🎯 What it is
Wise Council MCP is an MCP server that routes one question through multiple **methodology-based perspectives** instead of forcing a single synthetic answer.
The emphasis is on structured disagreement:
- different frameworks inspect the same question
- each perspective states a position and reasoning
- the final output preserves agreement, disagreement and open questions
- the user keeps the final decision
---
## 🔄 How it works
```mermaid
flowchart LR
Q[Your question] --> R[Router]
R --> D[Work / Product / Technology Desk]
R --> N[Life / Reflection Desk]
D --> P[Selected Perspectives]
N --> P
P --> M[Meeting Notes]
M --> O[Agreement · Disagreement · Questions · Actions]
```
The project intentionally avoids collapsing every perspective into one “perfect answer”. The disagreement is part of the output.
---
## ⚡ Quick Start
```bash
git clone https://github.com/Dream22180971/wise-council-mcp.git
cd wise-council-mcp
npm install
npm run build
npm start
```
For development:
```bash
npm run dev
```
---
## 🧩 Design
The server separates three responsibilities:
| Layer | Role |
|---|---|
| Router | decide which desk / perspective set fits the question |
| Council | generate independent structured viewpoints |
| Minutes | summarize agreements, disagreements, unresolved questions and actions |
A perspective is treated as a reasoning method, not as an attempt to impersonate a real person.
---
## 📋 Output Contract
A useful response format is:
```text
[Position]
What this perspective would prioritize.
[Reasoning]
Why it sees the problem this way.
[Question]
What it would ask before deciding.
```
Meeting notes then preserve:
- common ground
- unresolved disagreements
- assumptions
- questions for the user
- concrete next actions
---
## 🧠 MCP Use Cases
- product trade-offs
- architecture discussions
- career decisions
- writing critique
- risk review
- decision pre-mortems
- “what am I missing?” prompts
The tool is designed to add perspectives, not make decisions on behalf of the user.
---
## 🛠 Project Commands
```bash
npm run dev
npm run build
npm start
```
---
## 🗺 Roadmap
- [x] MCP server skeleton
- [x] Question routing
- [x] Multiple perspective sets
- [x] Structured meeting notes
- [ ] Configurable perspective packs
- [ ] User-defined councils
- [ ] Better trace / observability
- [ ] Evaluation set for routing quality
- [ ] More MCP client examples
---
## 🤝 Contributing
Useful contributions include new methodology packs, routing tests, output-contract improvements and MCP client examples.
---
## 📄 License
MIT
<div align="center">
**One question. Several frameworks. The decision stays yours.**
</div>
TDQS
Scored across 6 tools
Consult is clearly distinct as a single-expert tool, but review, critique, brainstorm, and roundtable all involve multiple experts generating input, and the boundaries between comprehensive review, critical critique, and divergent brainstorming can be blurry. Descriptions help, but some tools could be confused depending on the user's intent.
All tool names are single lowercase words, which gives a consistent visual and stylistic pattern. Most are verbs (consult, review, critique, suggest, brainstorm), though 'roundtable' is a noun used as an action, so it is a minor deviation.
Six tools is well-scoped for an advisory/council server. Each tool represents a distinct consultation mode without bloating the surface, and the count feels appropriate for the domain.
The server covers single-expert advice, structured group discussion, multi-angle review, brainstorming, critique, and scenario-based suggestions, which is a fairly complete advisory toolkit. A minor gap is the lack of tools for managing or listing available experts, but users can still interact without it.