Skip to content

GEMINI.md

Latest commit

 

History

History
87 lines (67 loc) · 3.49 KB

GEMINI.md

File metadata and controls

87 lines (67 loc) · 3.49 KB

Adobe MCP Controller

This project is a comprehensive Model Context Protocol (MCP) server system designed to control Adobe Creative Suite applications (Photoshop, Premiere Pro, Illustrator, InDesign, Acrobat) via AI agents. It enables natural language interaction with complex creative software.

Project Overview

The project is structured around a 3-tier architecture to bridge the gap between external AI agents and local Adobe desktop applications:

  1. MCP Servers (Python): act as the interface for AI agents (like Gemini or Claude). They translate natural language intent into structured commands.
  2. Proxy Server (Node.js/Python): Manages WebSocket connections to the Adobe applications, handling command routing and event listening.
  3. UXP Plugins (JavaScript): specialized plugins installed within each Adobe application that execute the actual commands via the Adobe API.

A parallel Rust implementation (adobe-mcp-rs) exists, specifically targeting Acrobat and potentially offering a higher-performance bridge in the future.

Key Directories

  • adobe-mcp-unified/: The main Python-based implementation.
    • adobe_mcp/: Source code for the MCP servers (photoshop, premiere, illustrator, indesign).
    • proxy-server/: Node.js WebSocket proxy (bridging MCP to UXP).
    • uxp-plugins/: Source code for the Adobe UXP plugins.
    • tests/: Python test suite.
  • adobe-mcp-rs/: Rust workspace for native implementations (currently focused on Acrobat).
  • scripts/: Global utility scripts (e.g., setup-adobe-agents.sh).

Setup and Installation

Prerequisites

  • Python 3.10+
  • Node.js 18+
  • Rust (cargo) - if working on adobe-mcp-rs
  • Adobe Creative Cloud Apps (Photoshop v26+, Premiere v25+, etc.)
  • Adobe UXP Developer Tools

Unified Python Setup (Primary)

Navigate to adobe-mcp-unified and use the provided scripts:

  • Windows: 
    .\install.ps1
    Or manually: pip install -e . and npm install in proxy-server.

Running the Servers

1. Start the Proxy

The proxy is required for communication with UXP plugins.

# From adobe-mcp-unified/
adobe-proxy
# OR
node proxy-server/proxy.js

2. Start MCP Servers

Each application has its own MCP entry point:

adobe-photoshop
adobe-premiere
adobe-illustrator
adobe-indesign

Note: Illustrator implementation currently includes a built-in proxy.

3. Load UXP Plugins

Use Adobe UXP Developer Tools to load the plugin manifest from adobe-mcp-unified/uxp-plugins/{app}/manifest.json into the target application.

Development Workflows

Testing

  • Run all tests (Windows): .\adobe-mcp-unified\run-tests.ps1
  • Python Tests: pytest inside adobe-mcp-unified

Multi-Agent Development

This project utilizes a multi-agent development strategy (detailed in MULTI_AGENT_PLAN.md). Key concepts:

  • Specialized Agents: Work is divided among agents like "UXP Research", "Premiere Dev", and "Infrastructure".
  • Worktrees: Development often happens in dedicated git worktrees (e.g., ../adobe-mcp-premiere).
  • Coordination: A central plan tracks task distribution.

Rust Development

For adobe-mcp-rs:

  • Build: cargo build
  • Test: cargo test

Important Configuration Files

  • adobe-mcp-unified/pyproject.toml: Python dependencies and entry points.
  • adobe-mcp-rs/Cargo.toml: Rust workspace members.
  • MULTI_AGENT_PLAN.md: Strategic roadmap for the multi-agent development process.