Skip to content

MCP Prompts

MCP Prompts are reusable, parameterized templates that clients can present to users. mcp-devtools includes four curated development prompts.

debug_error

Systematically debug an error using mcp-devtools tools.

ArgumentRequiredDescription
error_messageyesThe error message or stack trace
file_pathnoPath to the file where the error occurs

What it does: Guides the AI through a structured debugging workflow — reading the file (if provided), searching for related code patterns, checking logs, and proposing a fix.

code_review

Review a file for bugs, security issues, and code quality.

ArgumentRequiredDescription
file_pathyesPath to the file to review

What it does: Instructs the AI to read the file, check for error handling, input validation, security issues, and find related tests.

explore_codebase

Explore and understand a project's structure and conventions.

ArgumentRequiredDescription
focus_areanoArea to focus on (e.g., "database", "auth", "api")

What it does: Guides the AI to list directories, read key files (README, package.json, config), and summarize the architecture.

refactor_function

Refactor a function for readability, performance, or testability.

ArgumentRequiredDescription
file_pathyesPath to the file containing the function
function_nameyesName of the function to refactor

What it does: Instructs the AI to read the function, find usages and tests, analyze complexity, and propose a refactored version.

Usage

In compatible MCP clients, prompts appear as selectable templates. For example, in Claude Desktop you might select the debug_error prompt, provide the error message, and the AI will follow the structured debugging workflow using mcp-devtools tools.

Released under the MIT License.