mcpField
M
MCP Perforce Server

MCP Perforce Server

Local
github.com/Cocoon-AIVersion Control29Updated 45w ago

About MCP Perforce Server

MCP Perforce Server is an MCP server that wraps Perforce (P4) command-line operations for clients such as Claude Desktop. It is aimed at developers who use Perforce for version control and want commands that avoid interactive prompts. Per-project settings are read from .p4config files, and responses are returned in a structured form.

Key features

  1. 01Non-interactive execution of P4 commands
  2. 02Structured, parseable command output
  3. 03Per-project settings via .p4config files
  4. 04Common operations such as add, edit, sync and submit
  5. 05Changelist and stream management

Use cases

  1. 01Letting an AI assistant review pending workspace changes
  2. 02Creating and submitting a changelist from natural language requests
  3. 03Switching between projects that use different Perforce servers

MCP support

This is an MCP server that exposes P4 operations such as p4_status, p4_sync and p4_changelist_submit as tools to MCP clients.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"perforce": {
"command": "npx",
"args": [
"-y",
"@cocoon-ai/mcp-perforce"
],
"env": {
"P4CONFIG": "<P4CONFIG>"
}
}
}
}
Run commandnpx -y @cocoon-ai/mcp-perforce
Environment variables to set
P4CONFIG

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npm@cocoon-ai/mcp-perforce

Tools

25

Saved tool list · updated 3w ago

ParameterTypeRequiredDescription
filesarrayYesFiles to add (can use wildcards)
pathstringNoWorking directory path (optional)

Details

From the project README

MCP Perforce Server

A Model Context Protocol (MCP) server that provides a clean interface for Perforce (P4) operations in Claude Desktop. This server wraps P4 commands to make them more reliable and easier for Claude to use, eliminating issues with interactive prompts and complex state management.

Features

  • Non-interactive operations: All commands are wrapped to avoid interactive prompts
  • Structured responses: Clean, parseable output instead of raw P4 output
  • Multi-project support: Automatically uses .p4config files for per-project settings
  • Common operations: Add, edit, delete, submit, revert, sync, and more
  • Changelist management: Create and submit changelists with explicit parameters
  • Error handling: Graceful error handling with clear error messages

Installation

Prerequisites
  • Node.js 18+ installed
  • Perforce command-line client (p4) installed and in your PATH
  • .p4config files in your project directories (recommended)
Quick Install with Claude Code
# Install the package globally
npm install -g @cocoon-ai/mcp-perforce

# Add to Claude Code
claude mcp add perforce @cocoon-ai/mcp-perforce

That's it! Claude Code will automatically configure the server for you.

Manual Installation
Via NPM (when published)
npm install -g @cocoon-ai/mcp-perforce
Install from Source
git clone https://github.com/Cocoon-AI/mcp-perforce.git
cd mcp-perforce
npm install
npm run build
npm link  # Makes 'mcp-perforce' available globally

Configuration

Basic Setup

Add the server to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/claude/claude_desktop_config.json

{
  "mcpServers": {
    "perforce": {
      "command": "npx",
      "args": ["-y", "@cocoon-ai/mcp-perforce"],
      "env": {
        "P4CONFIG": ".p4config"
      }
    }
  }
}
Setting Up .p4config Files

The server uses Perforce's built-in P4CONFIG mechanism to automatically switch between different Perforce servers based on your current directory. Create a .p4config file in each project root:

# ~/projects/gamedev/.p4config
P4PORT=perforce-game.company.com:1666
P4CLIENT=gamedev-workspace
P4USER=your-username

# ~/projects/web/.p4config
P4PORT=perforce-web.company.com:1666
P4CLIENT=web-workspace
P4USER=your-username

Now the MCP server will automatically use the correct Perforce settings based on which project directory you're working in!

Available Commands

Once configured, you can ask Claude to use these P4 operations. All commands will automatically use the .p4config settings from your current project directory.

Basic File Operations
  • p4_status: Check workspace status and pending changes

    "Show me my pending P4 changes"
    "Check P4 status in the gamedev directory"
    
  • p4_add: Add files to Perforce

    "Add all .js files in the src directory to Perforce"
    
  • p4_edit: Open files for edit

    "Open config.json for editing in Perforce"
    
  • p4_delete: Mark files for deletion

    "Delete the old_module.py file from Perforce"
    
  • p4_sync: Sync files from depot

    "Sync all files in the project"
    "Force sync the src directory"
    
  • p4_revert: Revert files or entire changelists

    "Revert all files in changelist 12345"
    "Revert changes to config.json"
    
  • p4_diff: Show differences for files

    "Show me the diff for all my open files"
    
Changelist Operations
  • p4_changelist_create: Create a new changelist

    "Create a new changelist with description 'Fix login bug'"
    
  • p4_changelist_submit: Submit a changelist

    "Submit changelist 12345"
    
  • p4_move_to_changelist: Move files between changelists

    "Move all my open files to changelist 12345"
    
Stream Operations
  • p4_stream_list: List streams in a depot

    "List all streams in //depot"
    "Show me development streams matching 'feature'"
    
  • p4_stream_info: Get detailed stream information

    "Show me details about //depot/main stream"
    
  • p4_stream_switch: Switch workspace to a different stream

    "Switch to the //depot/dev stream"
    "Force switch to //depot/release-2.0"
    
  • p4_stream_create: Create a new stream

    "Create a development stream //depot/feature-xyz from //depot/main"
    
  • p4_stream_edit: Edit stream specification

    "Edit the //depot/feature-xyz stream spec"
    
  • p4_stream_graph: Show stream hierarchy

    "Show the stream hierarchy for //depot"
    
Client/Workspace Operations
  • p4_client_list: List all clients/workspaces

    "List all my Perforce workspaces"
    "Show clients for user jsmith"
    
  • p4_client_info: Get client/workspace details

    "Show me details about my current workspace"
    "Show info for client gamedev-workspace"
    
  • p4_client_create: Create a new client/workspace

    "Create a new workspace called dev-feature in /home/user/p4/feature"
    "Create a stream client for //depot/main-stream"
    
  • p4_client_edit: Edit client specification

    "Edit the view mappings for client dev-workspace"
    
  • p4_client_delete: Delete a client/workspace

    "Delete the old-feature workspace"
    "Force delete the broken-client workspace"
    
  • p4_client_switch: Switch to a different client

    "Switch to the production-client workspace"
    
Information Operations
  • p4_info: Show current Perforce configuration

    "Show me which P4 server and workspace I'm using"
    
  • mcp_perforce_version: Show MCP Perforce server version

    "What version of mcp-perforce is running?"
    

Example Usage in Claude

Here are some example conversations with Claude:

Example 1: Creating and submitting a change

You: "I've modified server.js and config.json. Create a changelist for these fixes"
Claude: I'll help you create a changelist for your changes. Let me first check the status...
[Uses p4_status, p4_changelist_create, p4_submit]

Example 2: Syncing and reviewing changes

You: "Sync the latest changes and show me what files I have open"
Claude: I'll sync your workspace and check your open files...
[Uses p4_sync, p4_status]

Troubleshooting

Server not appearing in Claude
  1. Restart Claude Desktop after modifying the configuration
  2. Check that the configuration file is valid JSON
  3. Verify the MCP server is installed: npm list -g @cocoon-ai/mcp-perforce
Perforce authentication errors
  1. Check your .p4config file exists and has the correct settings
  2. Test your connection outside Claude: p4 info
  3. Make sure you're logged in: p4 login
  4. Verify P4CONFIG is being recognized: p4 set P4CONFIG
Wrong Perforce server being used
  1. Check which directory you're in - the server uses .p4config from the current or parent directories
  2. Run p4 info to see which config is being used
  3. Use the p4_info command in Claude to debug: "Show me my P4 configuration"
Command not working as expected
  1. Check Claude's response for error messages
  2. Verify the P4 command works in your terminal from the same directory
  3. Enable debug logging (see below)

Debug Logging

To enable debug output, add to your configuration:

{
  "mcpServers": {
    "perforce": {
      "command": "npx",
      "args": ["-y", "@cocoon-ai/mcp-perforce"],
      "env": {
        "P4CONFIG": ".p4config",
        "DEBUG": "mcp:*"
      }
    }
  }
}

Development

Building from source
git clone https://github.com/Cocoon-AI/mcp-perforce.git
cd mcp-perforce
npm install
npm run build
Running tests
npm test
Adding new commands
  1. Add the tool definition in the appropriate file under src/tools/
  2. Add the handler function in the corresponding file under src/handlers/
  3. Update the switch statement in src/handlers/index.ts
  4. Follow the existing pattern for parameter validation and error handling

Contributing

Contributions are welcome! Please:

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/new-command)
  3. Commit your changes (git commit -am 'Add new P4 command')
  4. Push to the branch (git push origin feature/new-command)
  5. Create a Pull Request

License

MIT License - see LICENSE file for details

Acknowledgments

Support


Made with ❤️ for AIs struggling with P4 command-line operations

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.