Installing And Activating Codescene MCP

Installing and Activating CodeScene MCP

Overview

Use this skill when the task is to get the CodeScene MCP Server installed and registered with an AI assistant. The workflow has three parts: install the server, register it in the assistant, and copy the agent guidance files into the repository.

For configuring the server after installation (OAuth login, account ID, PAT, on-prem URLs, default project, SSL), use the configuring-codescene-mcp skill instead.

When to Use

• The user wants to install CodeScene MCP for the first time.
• The user has the binary or package installed but has not registered it in an AI assistant.
• The user wants a quick setup path for VS Code, GitHub Copilot, Claude Code, Docker, Homebrew, or Windows.

Do not use this skill for configuring tokens, URLs, or other server options. Use configuring-codescene-mcp for that. Do not use this skill for refactoring, safeguards, or technical debt prioritization.

Quick Reference

• Install using the method that matches the environment: npx or npm, Homebrew, Windows installer script, manual download, or Docker.
• Register the server in the AI assistant so it launches cs-mcp.
• Copy the appropriate agent guidance file into the repository: docs/AGENTS-full.md for CodeScene Core users, or docs/AGENTS-standalone.md for standalone license users. For Amazon Q, copy .amazonq/rules instead.
• Copy any relevant skills from the CodeScene MCP skills catalogue at https://github.com/codescene-oss/codescene-mcp-server/tree/main/skills.
• After installation, authenticate with OAuth via the login MCP prompt (slash command), by asking the assistant to log in (login tool), or with CodeScene: Sign In in the VS Code extension. Sign out with logout / CodeScene: Sign Out. Use a PAT only for CI/headless environments.

Implementation

1. Choose the installation method that fits the user environment:
   • npx @codescene/codehealth-mcp or npm install -g @codescene/codehealth-mcp
   • brew tap codescene-oss/codescene-mcp-server https://github.com/codescene-oss/codescene-mcp-server && brew trust codescene-oss/codescene-mcp-server && brew install cs-mcp
   • Windows PowerShell installer
   • Manual binary download
   • Docker image
2. Verify that the cs-mcp command or container entrypoint is available.
3. Register the server in the AI assistant:
   • For VS Code or GitHub Copilot, install the CodeScene CodeHealth MCP extension, or add a codescene server entry in settings.json / .vscode/mcp.json.
   • For Claude Code, install the plugin (/plugin install codescene@codescene) or add the MCP with claude mcp add.
4. Authenticate:
   • Interactive (recommended): Use the login MCP prompt (slash command), ask the assistant to log in (calls login and opens a browser), or run CodeScene: Sign In in the VS Code extension. For multi-account Cloud, use switch_account (or CodeScene: Switch Account) after login when needed. For on-prem, set onprem_url first.
   • CI / headless: Set a Personal Access Token via set_config or CS_ACCESS_TOKEN. See the configuring-codescene-mcp skill.
5. Copy the appropriate agent guidance file to the repository: docs/AGENTS-full.md for CodeScene Core users, or docs/AGENTS-standalone.md for standalone license users. Rename it to AGENTS.md in the target repository so the agent picks it up automatically.
6. If the user is on Amazon Q, copy .amazonq/rules instead of the AGENTS file.
7. Copy any relevant skills from https://github.com/codescene-oss/codescene-mcp-server/tree/main/skills for safeguarding, refactoring, prioritization, or business-case workflows.

Common Mistakes

• Installing the server but never registering it with the assistant.
• Copying the setup files into the wrong repository.
• Skipping authentication after installation. For interactive use, call login; for CI, set a PAT.
• Pasting a PAT into client env when OAuth would work — a set CS_ACCESS_TOKEN blocks login.
• Manually editing environment variables when login / set_config is simpler and persistent.