Deep Productivity MCP Server

MCP (Model Context Protocol) server that connects Claude Code to the Deep Productivity Pipeline database. Lets you read products, write artifacts, update documents, and more — directly from Claude.

Prerequisites

• Node.js 20+

Setup

1. Install and build

cd mcp-server
npm install
npm run build

2. Configure Claude Code

The .mcp.json in the project root is already configured with the Supabase connection. When you open the project in Claude Code, it will automatically start the MCP server. No additional environment variables or secrets needed.

3. Restart Claude Code

After the initial build, restart Claude Code for it to pick up the MCP server. The tools will then be available in your conversation.

Operating Modes

The MCP server supports three modes:

| Mode | Config | Behavior | |------|--------|----------| | Dynamic | CONTEXT_FILE env var | Product selection persisted to file. Switch products mid-conversation via set_active_product. get_active_product and set_active_product tools available. | | Scoped | PRODUCT_ID env var | Locked to one product. No switching without restart. | | Unscoped | Neither set | product_id required on every tool call. |

Dynamic mode is recommended. Both dp-connect and the main project repo use it.

Available Tools

Product Selection (Dynamic mode only)

| Tool | Description | Parameters | |------|-------------|------------| | get_active_product | Get the currently active product, or list all available products if none is selected. Call this at the start of every conversation. | — | | set_active_product | Set the active product. Persists across conversations until changed. | product_id (required) |

Products

| Tool | Description | Parameters | |------|-------------|------------| | list_products | List all products in the pipeline. Supports filtering. | stage?, status? | | get_product | Get full product details including team members, owner info, and all metadata. | product_id (optional in dynamic/scoped mode) |

Documents & Phase Data

| Tool | Description | Parameters | |------|-------------|------------| | get_phase_data | Get the document content (section key/value pairs) for a product and stage. | product_id?, stage | | update_document | Update a section of the phase document. Creates the row if it doesn't exist (upsert). | product_id?, stage, section_key, content |

Artifacts

| Tool | Description | Parameters | |------|-------------|------------| | list_artifacts | List all artifacts for a product and stage. Returns metadata only (no HTML body). | product_id?, stage | | get_artifact | Get the full content of a single artifact including its HTML body. | artifact_id | | write_artifact | Create a new artifact. Supports types: analysis, testPlan, diagram, matrix, report, website. For "website" type, provide a complete standalone HTML document with inline CSS and base64 images. | product_id?, stage, type, title, html | | update_artifact | Update an existing artifact's title, content, or type. | artifact_id, title?, html?, type? | | delete_artifact | Permanently delete an artifact by its ID. | artifact_id |

Comments

| Tool | Description | Parameters | |------|-------------|------------| | list_comments | List all comments for a product and stage, with author names. | product_id?, stage | | add_comment | Add a comment to a product stage. | product_id?, stage, author_id, text |

Files & Links

| Tool | Description | Parameters | |------|-------------|------------| | list_files_and_links | List all uploaded files and reference links for a product and stage. | product_id?, stage | | upload_file | Upload a local file to a product stage. Stored in Supabase Storage. | product_id?, stage, file_path, file_type? |

Pipeline Stages

Products move through 7 stages:

1. IDEAS — Initial concept and brainstorming
2. MARKET_ANALYSIS — Market research, TAM/SAM/SOM, competitive landscape
3. PROTOTYPE — Design and prototyping
4. BUILD — Engineering and development
5. TEST — QA, testing, and validation
6. LAUNCH — Go-to-market and launch execution
7. GROWTH — Post-launch growth and optimization

Connecting External Projects

Use the dp-connect CLI to connect any project directory to the pipeline:

# One-time install
npm install -g @deep-productivity/deep-productivity-mcp

# Connect to a product (fuzzy, case-insensitive)
dp-connect DeepFlash

This writes two files:

• .mcp.json — MCP server config (dynamic mode with CONTEXT_FILE)
• .claude/product-context.json — Selected product

Restart Claude Code after first-time setup. After that, you can switch products mid-conversation by asking Claude — no restart needed.

Example Usage

Once connected, just ask Claude naturally:

• "List all products in the BUILD stage"
• "Show me the team and details for Fizley"
• "What does the IDEAS document say for AI Code Review Bot?"
• "Write a competitive analysis artifact for Compliance Automation"
• "Update the market_size section for Data Pipeline Orchestrator"
• "Switch to the DeepFlash product"

When a product is active:

• "Show me the BUILD phase document" — no product ID needed
• "Write an analysis artifact for the TEST stage" — auto-targets active product
• "List all comments on MARKET_ANALYSIS" — scoped automatically

Everything you write through these tools shows up immediately in the web app.