Skip to content

graemerycyk/openpmkit

BranchesTags

Repository files navigation

OpenPMKit

Your PM workflows, inside your AI assistant. 13 product management workflows — daily briefs, PRDs, sprint reviews, competitor reports, and more — available through three installation routes depending on your setup.

You provide context (Slack messages, Jira tickets, customer feedback), and OpenPMKit generates structured, evidence-grounded PM artifacts through your AI assistant.

Who This Is For

Product managers who use AI assistants (Claude Desktop, Cursor, Claude Cowork, Claude Code) and want structured, repeatable workflows instead of writing prompts from scratch every time.

Three Ways to Use OpenPMKit

Route Best For Setup
A. MCP Server Claude Desktop, Cursor, any MCP client pip install + config
B. Claude Code Plugin Claude Code Desktop & CLI /plugin install
C. Claude Cowork Plugin Claude Cowork Marketplace install

Anthropic PR: OpenPMKit's workflows (8 commands, 5 skills) have been submitted to Anthropic's official product-management plugin. If accepted, they'll be available to every Cowork and Claude Code user who installs that plugin.

What You Get

13 PM Workflows as Tools

# Tool What It Produces Category
1 daily_brief Morning synthesis of Slack, Jira, support, community Autonomous
2 meeting_prep Account context, talking points, risks for customer meetings Autonomous
3 voc_clustering Customer feedback clustered into themes with impact scores On-Demand
4 competitor_research Competitor changes, feature gaps, strategic implications On-Demand
5 roadmap_alignment Decision memo with options, trade-offs, recommendation On-Demand
6 prd_draft Evidence-backed PRD with goals, requirements, timeline On-Demand
7 sprint_review Sprint summary, metrics dashboard, demo highlights Autonomous
8 prototype_generation Standalone interactive HTML prototype from a PRD On-Demand
9 release_notes Customer-facing release notes from completed work On-Demand
10 deck_content Slide content tailored for exec, team, customer, or stakeholder On-Demand
11 feature_ideation Raw ideas transformed into structured feature concepts Beta
12 one_pager 500-word executive summary from multiple inputs Beta
13 tldr 30-second Slack/email-ready bullet summary Beta

2 Utility Tools

  • pmkit_help — Lists every available tool with its required and optional fields
  • pmkit_workflow_details — Shows full field descriptions and examples for any workflow

Integrations

OpenPMKit connects to your PM stack through three methods:

MCP servers (pre-configured, connect automatically):

Integration What it provides
Slack Team messages, channel activity, stakeholder threads
Jira / Confluence (Atlassian) Sprint data, tickets, epics, existing docs
Linear Issues, projects, cycles
Asana Tasks, projects, milestones
Monday.com Boards, items, updates
Notion Pages, databases, meeting notes
Figma Design files, components, prototypes
Amplitude Usage metrics, funnels, cohorts
Intercom Support tickets, feature requests, conversations
Fireflies Meeting transcripts, action items
Microsoft 365 Email, calendar, documents

Claude built-in integrations (connect in Claude's settings):

Integration What it provides
Gmail Email threads, customer communications
Google Calendar Meeting schedules, attendee context
Google Drive Documents, spreadsheets, presentations

Supported via manual paste (no MCP endpoint yet):

Integration What it provides
Gong Call transcripts, revenue intelligence
Loom Video transcript insights
Zendesk Support ticket data
Zoom Meeting recordings, transcripts

AI-powered research (built-in skills, no setup needed):

Skill What it does
Competitive research Searches competitor sites, Product Hunt, G2, changelogs
Social monitoring Scans Reddit, Twitter/X, Hacker News, community forums
News tracking Tracks industry news, analyst reports, market developments

For full connector details including ~~category placeholders, see plugin/CONNECTORS.md.

How It Works

Each tool is self-documenting and conversational:

  1. Your AI assistant calls a tool (e.g. prd_draft)
  2. The MCP server checks if all required data is provided
  3. If anything is missing, it tells the assistant exactly what it needs — with descriptions and examples
  4. Once all data is there, it renders a complete structured prompt
  5. Your AI assistant uses that prompt to generate the artifact

You don't need to remember prompt formats. Just tell your assistant what you need in plain English.

Example Conversation

You: "I need a PRD for our new search filters feature. We've had 47 support tickets about search."

Claude: Calls prd_draft — sees it needs tenant_name, feature_name, and customer_evidence

Claude: "What's your company name?"

You: "Acme Corp"

Claude: Calls prd_draft with all fields, gets back a structured prompt, generates a full PRD

Installation

Option A: MCP Server (Claude Desktop / Cursor / any MCP client)

Prerequisites

  • Python 3.10 or newer — Check with python3 --version
  • pip (comes with Python) or uv (faster alternative)

Step 1: Clone the Repo

git clone https://github.com/graemerycyk/openpmkit.git
cd openpmkit

Step 2: Set Up Python Environment

# Create an isolated environment (keeps your system Python clean)
python3 -m venv .venv

# Activate it
source .venv/bin/activate        # macOS / Linux
# .venv\Scripts\activate          # Windows PowerShell
# .venv\Scripts\activate.bat      # Windows CMD

Step 3: Install

pip install -e .

That's it. The server is now installed and ready.

Verify It Works

pmkit-mcp

The server will start and wait for MCP client connections on stdio. Press Ctrl+C to stop. If it starts without errors, you're good.

Option B: Plugin (Claude Code Desktop & CLI)

No Python or server setup required. Install directly as a plugin with slash commands and skills:

Claude Code Desktop:

  1. Start a session, then type /plugin to open the plugin manager
  2. Go to the Marketplaces tab → click Add → enter graemerycyk/openpmkit
  3. Go to the Discover tab → find OpenPMKit → click Install

Claude Code CLI:

/plugin marketplace add graemerycyk/openpmkit
/plugin install openpmkit

Optionally specify a scope:

Scope Flag Effect
User (default) --scope user Available across all your projects
Project --scope project Shared with collaborators via git
Local --scope local This repo only, gitignored

All 15 slash commands will be available immediately. See plugin/README.md for the full command list.

Direct download: If the marketplace isn't working, download openpmkit.zip from this repo and install manually:

# Claude Code CLI
/plugin install /path/to/openpmkit.zip

# Or unzip to your plugins directory
unzip openpmkit.zip -d ~/.claude/plugins/

Option C: Plugin (Claude Cowork)

In the Cowork plugins sidebar, click Add marketplace from GitHub and enter:

graemerycyk/openpmkit

Or click Add marketplace by URL and enter:

https://github.com/graemerycyk/openpmkit.git

Then click Sync and install OpenPMKit from the marketplace list.

Note: Cowork plugin marketplace support is still in preview.

Standalone Skills (any Claude environment)

You can also copy individual workflow skills directly — no plugin install or server needed:

git clone https://github.com/graemerycyk/openpmkit.git
cp -r openpmkit/plugin/skills/* ~/.claude/skills/

Skills auto-activate when Claude detects a relevant task, or invoke directly with /skill-name.

Connect to Your AI Assistant (MCP Server)

The sections below apply to Option A (MCP Server) only. Options B and C handle connections automatically.

Claude Desktop (Recommended)

  1. Open Claude Desktop settings
  2. Go to Developer > Edit Config
  3. Add the PM Kit server:

macOS — edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "pmkit": {
      "command": "/full/path/to/openpmkit/.venv/bin/pmkit-mcp"
    }
  }
}

Windows — edit %APPDATA%\Claude\claude_desktop_config.json:

{
  "mcpServers": {
    "pmkit": {
      "command": "C:\\full\\path\\to\\openpmkit\\.venv\\Scripts\\pmkit-mcp.exe"
    }
  }
}
  1. Restart Claude Desktop
  2. You should see PM Kit tools in the tools list (hammer icon)

Using uv instead of pip?

{
  "mcpServers": {
    "pmkit": {
      "command": "uv",
      "args": ["run", "--directory", "/full/path/to/openpmkit", "pmkit-mcp"]
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "pmkit": {
      "command": "/full/path/to/openpmkit/.venv/bin/pmkit-mcp"
    }
  }
}

Claude Code (CLI)

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "pmkit": {
      "command": "/full/path/to/openpmkit/.venv/bin/pmkit-mcp"
    }
  }
}

Other MCP Clients

Any client that supports the MCP stdio transport can connect. Point it at the pmkit-mcp binary in your .venv/bin/ directory.

Usage

Once connected, just talk to your AI assistant naturally. Here are some things you can ask:

Discover tools

"What PM tools do I have?"

"Show me what workflows are available"

Daily work

"Generate my daily brief — I'm Jane at Acme Corp, here's today's Slack activity: [paste]"

"Prep me for my QBR with Globex Corp tomorrow"

Strategic work

"Draft a PRD for our new search filters feature based on this customer feedback: [paste]"

"Create a roadmap alignment memo — we need to decide between AI search and SSO for Q1"

"Analyze this competitor intel and tell me where we have gaps: [paste]"

Sprint work

"Generate a sprint review for Sprint 42, here are the completed stories: [paste]"

"Write release notes for v2.4.0 from these Jira tickets: [paste]"

Communication

"Create exec deck content for our Q4 product update"

"Give me a TL;DR of this sprint update for Slack"

"Turn this 3-page report into a one-pager for the board"

Prototyping

"Generate an HTML prototype from this PRD: [paste]"

Workflow Reference

Autonomous Workflows

Designed for recurring use — daily standups, meeting prep, sprint ceremonies:

  • Daily Brief (daily_brief) — Synthesizes Slack messages, Jira updates, support tickets, and community activity into a morning brief with urgent items, sprint progress, customer signals, and recommended actions.

  • Meeting Prep (meeting_prep) — Compiles account history, open issues, health metrics, and Gong call insights into a prep pack with talking points, questions to ask, and risks/opportunities.

  • Sprint Review (sprint_review) — Generates a stakeholder-ready sprint summary with key accomplishments, velocity metrics, demo highlights, blockers/learnings, and next sprint preview.

On-Demand Workflows

Run these when you need a specific artifact:

  • VoC Clustering (voc_clustering) — Clusters feedback from support, calls, community, and NPS into themes with mention counts, representative quotes, trend analysis, and prioritized recommendations.

  • Competitor Research (competitor_research) — Tracks competitor product changes with feature gap analysis, strategic implications, and recommended responses.

  • Roadmap Alignment (roadmap_alignment) — Creates a decision memo with 2-3 options, evidence for each, trade-offs, a recommendation, and next steps. Formatted for executive review.

  • PRD Draft (prd_draft) — Drafts a full PRD: problem statement, goals, user stories, requirements, success criteria, assumptions, risks, and timeline. All grounded in customer evidence.

  • Prototype Generation (prototype_generation) — Outputs a standalone HTML file with embedded CSS and JavaScript. Interactive, responsive, and ready to open in a browser.

  • Release Notes (release_notes) — Translates Jira tickets into customer-friendly release notes with highlights, new features, improvements, bug fixes, and a "coming soon" preview.

  • Deck Content (deck_content) — Generates slide-by-slide content with headlines, bullets, key metrics, visual suggestions, and speaker notes. Adapts tone and depth to the audience (exec, team, customer, stakeholder).

Beta Workflows

Newer, still being refined:

  • Feature Ideation (feature_ideation) — Takes raw ideas, Slack threads, customer signals, and constraints and produces a structured feature concept with problem definition, solution options, assumptions to validate, and 2-week action plan.

  • One-Pager (one_pager) — Distills multiple documents into a 400-500 word executive summary with TL;DR, context, key findings, recommendations, and next steps.

  • TL;DR (tldr) — Creates a 3-5 bullet summary under 100 words, optimized for Slack and mobile. Includes emoji for scanning and a clear call-to-action.

Project Structure

openpmkit/
├── pmkit_mcp/                   # Route A: MCP server package
│   ├── __init__.py              # Version metadata
│   ├── __main__.py              # python -m pmkit_mcp entry point
│   ├── server.py                # MCP server — tool registration and handlers
│   ├── renderer.py              # Template rendering and field validation
│   └── workflows/
│       ├── __init__.py          # Public API
│       └── registry.py          # All 13 workflow definitions (source of truth)
├── plugin/                      # Routes B & C: Claude Code / Cowork plugin
│   ├── .claude-plugin/
│   │   └── plugin.json          # Plugin manifest
│   ├── commands/                # 15 slash command .md files
│   └── README.md                # Plugin installation and usage
├── .claude-plugin/
│   └── marketplace.json         # Plugin marketplace catalog
├── scripts/
│   └── build_plugin.py          # Generates command .md files from registry.py
├── tests/
│   ├── test_registry.py         # Workflow definition integrity
│   ├── test_renderer.py         # Prompt rendering correctness
│   └── test_server.py           # MCP tool call integration
├── pyproject.toml               # Package config, dependencies, tool settings
├── CLAUDE.md                    # Quick reference for AI coding assistants
├── AGENTS.md                    # Full developer and contributor guide
└── README.md                    # This file

Architecture

You (natural language)
 │
 ▼
AI Assistant (Claude Desktop / Cursor / Claude Code)
 │
 │ MCP protocol (stdio, JSON-RPC)
 │
 ▼
┌──────────────────────────────┐
│     PM Kit MCP Server        │
│                              │
│  pmkit_help ──────────────── │ ──→ Lists all 13 tools
│  pmkit_workflow_details ──── │ ──→ Field docs for one tool
│                              │
│  Tool Router                 │
│    │                         │
│    ├─ Schema validation      │ ──→ Rejects calls missing required fields
│    │                         │
│    ├─ Renderer               │ ──→ Substitutes {{placeholders}} with your data
│    │                         │
│    └─ Workflow Registry      │ ──→ 13 workflow definitions with prompts
│                              │
│  Returns: {system_prompt,    │
│            user_prompt}      │
└──────────────────────────────┘
 │
 ▼
AI Assistant processes the structured prompt → generates your artifact

The server is stateless (no database, no files, no sessions) and LLM-agnostic (it produces prompts — your AI assistant's model processes them). Works with Claude, GPT, Gemini, or any model behind your MCP client.

Troubleshooting

"Command not found: pmkit-mcp"

Make sure your virtual environment is activated:

source /path/to/openpmkit/.venv/bin/activate

Or use the full path in your MCP config:

{ "command": "/full/path/to/openpmkit/.venv/bin/pmkit-mcp" }

Claude Desktop doesn't show PM Kit tools

  1. Check the config file path is correct for your OS
  2. Make sure you used the full absolute path to pmkit-mcp (not a relative path)
  3. Restart Claude Desktop completely (quit and reopen)
  4. Check Claude Desktop logs for errors

"ModuleNotFoundError: No module named 'mcp'"

You need to install the package:

cd /path/to/openpmkit
source .venv/bin/activate
pip install -e .

Tools appear but don't work

Check the server runs standalone:

pmkit-mcp

If it starts without errors and waits for input, the server is fine. The issue is in the MCP client configuration.

Development

For contributors and developers who want to modify workflows or add new ones.

Setup

git clone https://github.com/graemerycyk/openpmkit.git
cd openpmkit
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Run Tests

pytest              # all 24 tests
pytest -v           # verbose output
pytest tests/test_server.py  # just server tests

Lint and Type Check

ruff check pmkit_mcp tests    # linting
mypy pmkit_mcp                # type checking

Adding a New Workflow

See AGENTS.md for the step-by-step guide.

License

MIT

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

2 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages