AI Coder Tools MCP Servers

This repository contains Model Context Protocol (MCP) server implementations that extend Roo Code capabilities with various tools. For detailed documentation of each tool's usage and examples, please visit our tools documentation.

Quick Navigation

flowchart LR
    Installation["📦 Installation<br><font size=2>Get started with Windows/Mac Installations</font>"] --- Docs["📄 Docs<br><font size=2>Explore docs, demos, and best practices</font>"] --- Teams["👥 Teams Channel<br><font size=2>Join discussions and get support</font>"] --- BugReport["🐛 Bug Report<br><font size=2>Report issues or suggest improvements</font>"]
    
    click Installation "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#installation-options-overview" "View installation options"
    click Docs "https://github.com/ai-microsoft/ai-coder-tools/blob/main/wiki/README.md" "Read documentation"
    click Teams "https://teams.microsoft.com/l/team/19%3A_xydaMH0gjMpJALaw6mV1F_t9GLE-ojvKxlRFL27hNk1%40thread.tacv2/conversations?groupId=8f030644-3a69-4f8a-8517-3e88214487b8&tenantId=72f988bf-86f1-41af-91ab-2d7cd011db47" "Join Teams channel"
    click BugReport "https://github.com/ai-microsoft/ai-coder-tools/issues" "Submit a bug report"
    
    style Installation fill:#e6f7ff,stroke:#0078d4,stroke-width:2px
    style Docs fill:#e6ffe6,stroke:#107c10,stroke-width:2px
    style Teams fill:#e8d6ff,stroke:#5c2d91,stroke-width:2px
    style BugReport fill:#fff0e6,stroke:#d83b01,stroke-width:2px
    
    linkStyle 0,1,2 stroke-width:0px

Installation Options Overview

Note: Access Installation Issues for known installation issues.

flowchart TD
    Start([AI Coder Tools]) --> UserType{User Type}
    
    %% Windows User Path
    UserType -->|Windows User| WinPrereq[Install PowerShell 7]
    WinPrereq --> WinAllInOne[All-In-One Installation]
    WinAllInOne --> WinConfig[MCP Server Configuration]
    WinConfig --> WinStartup[Startup Setting]
    
    %% macOS User Path
    UserType -->|macOS User| MacPrereq[Terminal + Homebrew]
    MacPrereq --> MacAllInOne[All-In-One Installation]
    MacAllInOne --> MacConfig[MCP Server Configuration]
    MacConfig --> MacStartup[Startup Setting]
    
    %% Contributors Path
    UserType -->|Contributors| DevPrereq[Node.js 20.18.3 + TypeScript]
    DevPrereq --> DevSetup[Clone Repository and Build]
    DevSetup --> DevOptions[Development Options]
    DevOptions --> DebugMode[Debug Mode]
    DevOptions --> ExpTools[Experimental Tools]
    
    class Start,UserType fill:#f9f,stroke:#333,stroke-width:2px
    class WinPrereq,MacPrereq,DevPrereq fill:#ddf,stroke:#333,stroke-width:1px
    class WinAllInOne,MacManual,DevSetup fill:#9e9,stroke:#333,stroke-width:1px
    class WinConfig,MacConfig,DevOptions fill:#fda,stroke:#333,stroke-width:1px
    class WinStartup,MacStartup fill:#fda,stroke:#333,stroke-width:1px
    class WinVideo,MacVideo fill:#d9f7be,stroke:#333,stroke-width:1px
    class DebugMode,ExpTools fill:#ddd,stroke:#333,stroke-width:1px

    %% Clickable links
    click WinPrereq "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#for-windows-users" "View Windows prerequisites"
    click WinAllInOne "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#all-in-one-installation" "View All-In-One installation steps"
    click WinConfig "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#mcp-server-configuration" "View MCP server configuration"
    click WinStartup "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#startup-setting" "View startup settings"
    click MacPrereq "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#for-macos-users" "View macOS prerequisites"
    click MacManual "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#manual-installation" "View manual installation steps"
    click MacConfig "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#mcp-server-configuration" "View MCP server configuration"
    click MacStartup "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#startup-setting" "View startup settings"
    click MacVideo "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#installation-video" "Watch installation video"
    click DevPrereq "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#prerequisites" "View contributor prerequisites"
    click DevSetup "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#development-setup" "View development setup steps"
    click DebugMode "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#debug-mode-installation" "View debug mode installation"
    click ExpTools "https://github.com/ai-microsoft/ai-coder-tools/blob/main/ts-server/README.md#enable-experimental-tools" "View experimental tools setup"

For Windows Users

System Prerequisites

• Install PowerShell 7 or later - One-line silent install:

winget install --id Microsoft.PowerShell --silent

• In Windows Search Toolbar, input pwsh to open PowerShell 7

All-In-One Installation

1. Watch the video for installation with AllInOne script: AllInOne Installation

2. Get AllInOne.ps1 in assets from latest release

NOTE: Just AllInOne.ps1 is enough and no need to download all assets

3. Run PowerShell 7 as Administrator and execute following command(aka: pwsh)

pwsh -ExecutionPolicy Bypass -File .\\AllInOne.ps1

• Support paramters:

  • For experimental tools, please enable with parameter -enableExperimentalTool.

• TroubleShooting

  • Refer to System Prerequisites if you don't have powershell 7
  • Refer to Installation Issues for known installation issues.
  • If no similar issue found, Welcome to open New issues !

4. Refer to section MCP Server Configuration to customize your mcpserver Env setting
5. Restart vs code to apply the changes

For MacOS Users

System Prerequisites

• Terminal access (built into MacOS)
• Homebrew (optional, but recommended for easier Node.js installation):

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

All-In-One Installation

1. Get AllInOne.sh in assets from latest release. Just download the script, no need to download all assets. You can also download it from source code if it's absent in the release package.

2. Run the script in Terminal:

chmod +x AllInOne.sh
./AllInOne.sh

3. Restart VS Code to apply the changes.

4. There is an issue enabling Github Copilot Claude 3.7 so this part is disabled in the script. Should be fixed soon.

5. Troubleshooting guide

   • AllInOne script - Failed to get latest Roo Cline release info

     • It's usually an auth issue. Please logout in your terminal with gh auth logout and rerun the script. When logging in, make sure to choose your microsoft account and authorize Microsoft AI org.

Manual Installation

1. Download latest vsix of MS Roo Code from release page and install it in VS Code.
2. Download the latest ai-coder-tools package from the releases page
3. Unzip the release package, inside the unzipped folder, also unzip the ai-coder-tools-{version}.tgz, find the setup.sh inside.
   • Make sure the line endings is LF instead of CRLF, you can open setup.sh with VSCode and save it with LF line endings.
4. Open Terminal and navigate to the unzipped folder
5. Make the setup script executable and run it:

chmod +x setup.sh
./setup.sh

The setup script will:

• Install Node.js v20.18.3 if needed
• Install the package globally
• Configure MCP settings automatically

MCP Server Configuration

When installation finished, the MCP settings file is located at:

Windows:

%APPDATA%\\Code\\User\\globalStorage\\microsoftai.ms-roo-cline\\settings\\cline_mcp_settings.json

you can put this path directly to Windows search toolbar to & edit it.

MacOS:

~/Library/Application Support/Code/User/globalStorage/microsoftai.ms-roo-cline/settings/cline_mcp_settings.json

The content would be like below. And please note the envs are optional. If not provided, you will need to provide the full ado urls for tool's input parameter.

        "AZURE_DEVOPS_ORG_URL": "https://dev.azure.com/your-org",
        "AZURE_DEVOPS_DEFAULT_PROJECT_NAME": "your-project",
        "AZURE_DEVOPS_DEFAULT_REPO_NAME": "your-repo"

NOTE: If you're using nvm to manage node package, please update %NPM_PREFIX% to the value as the below command output

$nvmCurrentVersion = node -v
Join-Path $env:NVM_HOME $nvmCurrentVersion

{
  "mcpServers": {
    "ado-tools": {
      "command": "node",
      "args": [
        "%NPM_PREFIX%/node_modules/ai-coder-tools/build/servers/ado_tools_server.js"
      ],
      "env": {
        "AZURE_DEVOPS_ORG_URL": "https://dev.azure.com/your-org",
        "AZURE_DEVOPS_DEFAULT_PROJECT_NAME": "your-project",
        "AZURE_DEVOPS_DEFAULT_REPO_NAME": "your-repo"
      },
      "disabled": false,
      "alwaysAllow": [],
      "autoApprove": []
    },
    "git-tools": {
      "command": "node",
      "args": [
        "%NPM_PREFIX%/node_modules/ai-coder-tools/build/servers/git_tools_server.js"
      ],
      "disabled": false,
      "alwaysAllow": [],
      "autoApprove": []
    },
    "kusto-tools": {
      "command": "node",
      "args": [
        "%NPM_PREFIX%/node_modules/ai-coder-tools/build/servers/kusto_tools_server.js"
      ],
      "env": {
        "KUSTO_DEFAULT_CLUSTER": "https://your-cluster.kusto.windows.net",
        "KUSTO_DEFAULT_DATABASE": "your-database"
      },
      "disabled": false,
      "alwaysAllow": [],
      "autoApprove": []
    },
    "search-tools": {
      "command": "node",
      "args": [
        "%NPM_PREFIX%/node_modules/ai-coder-tools/build/servers/search_tools_server.js"
      ],
      "disabled": false,
      "alwaysAllow": ["search_bing"],
      "autoApprove": []
    }
  }
}

Startup Setting

• In VS Code's left panel, you can find the tab to MS ROO Code startup page

• In MS ROO CODE startup page, select API Provider VS Code LM API

• If the Language Model list is unavailable, please log in to GitHub Copilot, switch to another provider, then switch back to VS Code LM API.

For Contributors

Prerequisites

• Node.js >= 20.18.1

  NOTE: Latest Node.js version (v22.14.0) may not work. Use v20.18.3.

• TypeScript
• Azure DevOps account (for ADO-related tools)

Development Setup

1. Clone the repository:

git clone https://github.com/ai-microsoft/ai-coder-tools.git
cd ai-coder-tools/ts-server

2. Install dependencies:

npm install

3. Build the server:

npm run build

Development Installation Options

Debug Mode Installation

For development and testing:

Windows:

pwsh -ExecutionPolicy Bypass -File .\\setup.ps1 -debug on

MacOS:

./setup.sh --debug on

Debug mode features:

• Requires package.json in the repository root or user-specified path
• Creates automatic backup of existing MCP settings
• Installs from source instead of global package
• Uses local build for server paths

To disable debug mode and restore original configuration:

Windows:

pwsh -ExecutionPolicy Bypass -File .\\setup.ps1 -debug off

MacOS:

./setup.sh --debug off

Enable Experimental Tools

To enable experimental features like context-entity-tools:

Windows:

pwsh -ExecutionPolicy Bypass -File .\\setup.ps1 -enableExperimentalTools

MacOS:

./setup.sh --enableExperimentalTools

Prerequisites for experimental tools:

• Ollama installation:
  • Windows: Visit https://ollama.ai/download
  • MacOS: brew install ollama
• Running Ollama service (starts automatically on macOS after installation)
• 'all-minilm' model installed (ollama pull all-minilm)

You can combine both options:

Windows:

pwsh -ExecutionPolicy Bypass -File .\\setup.ps1 -debug on -enableExperimentalTools

MacOS:

./setup.sh --debug on --enableExperimentalTools

Configuration Fields

• command: The command to run the server (usually "node")
• args: Array containing the path to the specific server implementation file
• env: Environment variables required by the server (varies by server type)
• disabled: Whether the server is disabled (default: false)
• alwaysAllow: Array of tool names that are always allowed without approval
• autoApprove: Array of tool names that can be auto-approved (default: empty)

For detailed instructions on creating new tools and contributing to the project, please see README_CONTRIBUTING.md.