FAQ: CodeScene Single-User MCP
Use this page for quick answers to the most common questions about CodeScene Single-User MCP. For step-by-step guidance, start with the CodeScene’s Single-User MCP Help Center collection. For deeper technical detail, use the GitHub docs.
What is CodeScene Single-User MCP?
CodeScene Single-User MCP is a standalone edition of the CodeScene MCP Server focused on local Code Health workflows. It gives your AI assistant direct access to Code Health so it can review code, safeguard AI-generated changes, support refactoring, and explain why Code Health matters.
Do I need a full CodeScene subscription to use it?
No. CodeScene Single-User MCP was created so you can use local Code Health analysis without subscribing to a full CodeScene subscription.
What does the Single-User edition include?
A Single-User access token enables local Code Health analysis tools. That includes scoring, review, and refactoring-related workflows. Project-level and API-dependent features are included in the full CodeScene subscription.
Which tools are part of the Single-User workflow?
The Single-User workflow covers local Code Health tools such as:
code_health_score
code_health_review
pre_commit_code_health_safeguard
analyze_change_set
code_health_refactoring_business_case
explain_code_health
explain_code_health_productivity
If ACE is configured with CS_ACE_ACCESS_TOKEN, code_health_auto_refactor is also available.
Do I need ACE to use CodeScene Single-User MCP?
No. ACE is optional. CodeScene MCP works without ACE, and ACE becomes available only when CS_ACE_ACCESS_TOKEN is set. ACE requires an additional license.
Do I need to share my code with CodeScene?
No. For local analysis, the CodeScene’s Single-User MCP server runs locally and processes analysis on your machine against your local repository. The GitHub FAQ states that no source code or analysis data is sent to cloud providers, LLM vendors, or external services as part of local analysis. The documented Docker setup also mounts the repository read-only.
Why do I need to copy AGENTS.md into my repository?
AGENTS.md tells AI tools how to use CodeScene MCP correctly. It contains rules that safeguard AI coding, prevent regressions, and keep agents aligned with Code Health metrics. If you use Amazon Q, copy .amazonq/rules instead.
Which installation methods are supported?
CodeScene MCP supports:
NPM / npxHomebrewWindows PowerShell installDockermanual download from GitHub ReleasesWhere do I get the Single-User access token?
See our article How to start your Single-User MCP free trial and get your access token.
Where should I look for help first?
Use this order:
Single-User MCP Help Center collection
Chatbot Eve in the Help Center
What can I use the MCP server for?
In short, the CodeScene MCP server gives AI coding assistants deterministic Code Health guidance so they can generate healthier code, refactor more safely, and avoid introducing technical debt. It connects AI tools to CodeScene’s CodeHealth analysis, helps safeguard AI-generated code, supports more targeted refactoring, simplifies code reviews with maintainability checks, and helps build a business case for refactoring through ROI-oriented workflows.
Does the MCP server run locally?
Yes. The CodeScene MCP server is designed to run in your local environment.
How does the MCP server keep my code private and secure?
The CodeScene MCP server runs fully locally. Code Health scoring, delta reviews, and business-case calculations are performed on your machine against your local repository, without sending source code or analysis data to cloud providers, LLM vendors, or external services.
Does the CodeHealth MCP Server work with tools like GitHub Copilot or Cursor?
Yes. The CodeScene MCP server works with AI coding assistants including GitHub Copilot, Cursor, Windsurf, and similar development environments. Because it follows CodeScene MCP, it also fits into broader AI-assisted workflows that support the protocol.
Does the MCP Server support agentic workflows?
Yes. The CodeScene MCP server is designed for agentic workflows and composable AI tooling rather than being tied to a single editor, assistant, or model.
Does the MCP server support multiple programming languages?
Yes. CodeScene supports 30+ programming languages, so the CodeScene MCP server can be used across polyglot codebases.
Can I use any LLM as the backbone for CodeScene’s MCP server?
The CodeScene MCP server can work with any model your AI assistant supports. When your assistant offers model selection, but we strongly recommend using a newer frontier model because those models tend to follow CodeScene’s MCP constraints and refactoring workflows more reliably.
How do I get started with the MCP server?
Start by setting your token as the CS_ACCESS_TOKEN environment variable. Then install the CodeScene MCP server through one of the supported installation methods, connect it to your AI assistant, and add AGENTS.md to your repository so the agent follows the intended workflow and safeguards.
Is there a free trial available?
Yes. The CodeScene MCP Server includes a 30 days free trial. After the trial ends, the subscription continues as a paid plan unless you cancel.
How much does the CodeHealth MCP Server cost?
The paid plan costs €8 or $9 per month. Customers in the United States are billed in USD, while customers in other regions are billed in EUR. Annual billing includes a 10% discount.
Can I cancel my subscription anytime?
Yes. You can cancel your subscription at any time, and access remains active until the end of the current billing period.
How does the CodeHealth MCP server safeguard AI-generated code?
The CodeScene MCP server helps prevent AI tools from introducing technical debt by surfacing maintainability issues such as high complexity, deep nesting, and low cohesion. It uses objective Code Health signals and a self-correcting loop so that when quality drops or risk increases, the AI is pushed to adjust the code and re-check it before moving forward.
How do I use the MCP server to refactor code and expand the AI-ready areas?
The typical flow is review, plan, refactor, and re-measure. The CodeScene MCP server gives the AI assistant deterministic guidance through Code Health reviews, helps identify the specific maintainability issues to address, and supports refactoring in smaller measurable steps so unhealthy legacy code becomes easier and safer for AI-assisted work.
How does CodeScene measure AI-ready code?
CodeScene uses CodeHealth as a proxy for AI readiness. It aggregates multiple structural maintainability factors to measure how easy code is to understand, modify, and evolve, with healthier code being safer and more reliable for AI agents to work on.
What is the magic number for AI-ready code?
The target for AI-ready code is a Code Health score of at least 9.5, ideally 10.0. Code below that may need refactoring first so AI can work with it more safely and effectively.