KeystoneMCP

Overview

KeystoneMCP is a local-first production-context and project-management MCP server that provides persistent memory for your AI assistants. It acts as the authoritative context layer across sessions, allowing your AI to understand the nuances of complex creative projects without replacing your existing tools.

How it Works

Instead of replacing your creative software, KeystoneMCP works in tandem with other MCP servers. Your AI client consults KeystoneMCP for project state while connecting directly to creative tools like Houdini, Maya, Nuke, Blender, and ComfyUI. It records completed work back into a persistent production state.

Key Features

• Persistent Project State: Maintain context across multiple AI sessions.
• Structured Production Graph: Manage assets, production units, tasks, versions, and dependencies.
i
• Version Tracking: Track current and approved versions of project elements.
• Workflow Templates: Access recommended next actions and preflight checks.
i
• Downstream Impact Analysis: Understand what could be affected when a change is made.
• Naming & Path Rules: Establish clear rules for source, work, review, publication, and delivery files.
• Profile-Based Creation: Create project folder structures based on Film & Animation, Game Development, or Creative Production profiles.

Supported Workflows

KeystoneMCP is designed for various creative production workflows including:

1. Film & Animation
2. Game Development
3. Graphic Design
4. Photography
5. Marketing and other creative-production pipelines.

Technical Requirements

To get started, you will need the following:

• Python 3.11 or newer
• Node.js 18 or newer (LTS recommended)
• A compatible AI client with MCP support (e.g., Claude)
• macOS or Linux (Windows users require WSL)

Installation & Setup

Follow these steps to set up KeystoneMCP:

1. Open Terminal and navigate to the extracted folder.
2. Run ./keystone setup to configure your project storage location.
3. Run ./keystone start to launch the server.
4. Use the Connect MCP button in the Web UI (http://127.0.0.1:8765) to obtain configuration for your AI client.