# MCP Product Decision Copilot

An MCP server that gives Claude product intelligence tools to make **Ship / Delay / Kill** recommendations on feature requests. Built for TaskFlow, a B2B SaaS project management platform.

## What It Does

Three tools provide Claude with structured product data:

| Tool | Input | Returns |
|------|-------|---------|
| `get_metrics(feature_area)` | `collaboration`, `workflows`, `reporting`, `integrations` | Usage, retention, support tickets, revenue, user feedback |
| `get_roadmap(quarter)` | `2026-Q1` | Team capacity, commitments, tech debt, dependencies, risks |
| `get_okrs(quarter)` | `2026-Q1` | Company and product team OKRs, strategic themes |

Claude calls all three tools, then produces a structured recommendation with decision, confidence level, reasoning, risks, and next steps.

## Prerequisites

- [uv](https://docs.astral.sh/uv/) (Python package manager)
- [Claude Desktop](https://claude.ai/download)

## Setup

1. Clone the repo and install dependencies:

```bash
cd "MCP Copilot"
uv sync
```

2. Add the server to Claude Desktop. Open **Claude > Settings > Developer > Edit Config** and add:

```json
{
  "mcpServers": {
    "product-copilot": {
      "command": "uv",
      "args": [
        "run",
        "--directory", "/ABSOLUTE/PATH/TO/MCP Copilot",
        "python", "server.py"
      ]
    }
  }
}
```

Replace `/ABSOLUTE/PATH/TO/MCP Copilot` with the actual path to this project.

3. Restart Claude Desktop (Cmd+Q, then relaunch). You should see 3 tools in the tools menu.

## Demo Scenarios

Try these prompts in Claude Desktop to see the copilot in action:

**Scenario A — Ship:** "We're getting requests from enterprise prospects for real-time collaborative editing. Should we build this for Q1?"

**Scenario B — Delay:** "Power users want advanced workflow automation — if/else branching, loops, scheduled triggers. Should we prioritize this for Q1?"

**Scenario C — Kill:** "I've been thinking about building a public API marketplace for TaskFlow. Should we invest in this for Q1?"

**Scenario D — Tradeoff:** "There's a surge of interest in AI-powered task prioritization. Should we build this for Q1?"

## Project Structure

```
MCP Copilot/
  server.py              # MCP server (3 tools)
  pyproject.toml         # Python dependencies
  data/
    metrics/             # Per-feature-area product metrics
    roadmap/             # Per-quarter engineering roadmap
    okrs/                # Per-quarter company and team OKRs
  prompts/
    system_prompt.md     # Decision framework for Claude
    demo_prompts.md      # Full demo prompt text
```