MCP Creator Growth

What SunflowersLwtech/mcp_creator_growth MCP server does

SunflowersLwtech/mcp_creator_growth MCP server, published in the repository as Covate, provides a learning sidecar for AI coding assistants. It addresses two related workflows: helping a developer understand AI-generated changes and retaining debugging knowledge for a specific project.

The learning workflow uses interactive quizzes about recent code modifications. Questions can focus on logic, security, performance, architecture, or syntax. The debugging workflow stores error context, root-cause analysis, and successful fixes, then makes those records searchable later. A separate reference tool returns programming terms from supported technical domains and tracks terms that have already been shown.

How it works

The server exposes four MCP tools. learning_session opens a WebUI quiz and blocks until the user completes it. Its inputs can include a project directory, an agent-generated change summary, optional five-Why reasoning, quiz questions, focus areas, and a timeout. On completion, it returns a status indicating that generation should halt.

debug_search accepts an error message or description and searches the project’s debugging history. Results can be narrowed by error type or tags, with a configurable result limit. debug_record stores a fixed bug’s context, cause, solution, project directory, and optional tags. The README describes search and recording as silent operations that agents can invoke around error handling and fixes. term_get returns one to five terms from domains such as algorithms, testing, security, databases, DevOps, and version control.

Setup and configuration

SunflowersLwtech/mcp_creator_growth MCP server requires Python 3.11+ or uv. The documented installer clones the project into ~/covate, creates an environment, installs dependencies, and prints an IDE configuration command. Manual setup uses either uv with an editable install or a standard Python virtual environment. Docker is also supported, including the published sunflowerslwtech/covate:latest image.

The README provides Claude Code configuration through the claude mcp add command or a ~/.claude.json entry. Docker documentation also covers Claude Desktop integration. The core server runs locally without an account. An optional hosted learning ledger can sync sessions to Covate after GitHub sign-in, using COVATE_SYNC_URL and COVATE_SYNC_TOKEN; this is separate from the local MCP functionality.

Tools and capabilities

• Run blocking quizzes based on recent code changes.
• Focus quizzes on logic, security, performance, architecture, or syntax.
• Search project-specific debugging history.
• Record error context and verified solutions.
• Retrieve programming terms across ten listed domains.
• Track displayed terms to reduce repetition.

Limitations and notes

The quiz tool blocks until the user finishes the learning interaction, so it is not suited to workflows that cannot pause generation. Debugging memory is project-specific and depends on agents recording useful context and solutions. The README claims token optimization and multilingual support in the repository description, but the excerpt does not explain how those capabilities work or which languages are covered.

The hosted ledger is optional and free, but it is not required for local operation. Progress-over-time charts, spaced-repetition plans, and team accounts are explicitly described as unavailable. The repository is MIT licensed, and the provided setup targets local execution rather than a required paid API or hosted service.