bullet

bullet-mcp is an MCP server that validates and improves bullet point lists using evidence-based cognitive psychology and UX principles. It scores bullet lists against seven rules such as list length, hierarchy, line length, serial position, parallel structure, first words, and formatting, then provides a letter grade, actionable feedback, and citations for each validation rule. The tool supports both single-section inputs and sectioned, multi-chapter documents, making it suitable for quick lists as well as longer documents with structured sections. To run it via npx, you can invoke the server through the CLI and feed it a JSON payload describing the bullets and their context (e.g., document, presentation, or reference). The output includes an overall score, grade, and suggested improvements to help you optimize readability and recall.

Prerequisites:

• Node.js and npm installed on your machine

Installation options:

1. Local project install (recommended for development):

   • Run: npm install bullet-mcp
   • Use the tool by invoking the package CLI or requiring it in your code as documented in the package's README.

2. Global install (quick start):

   • Run: npm install -g bullet-mcp
   • Then run the bullet-mcp command directly from your shell as described in the package documentation.

3. Run via npx (no installation required):

   • Run: npx bullet-mcp
   • This will download and execute the latest version from npm.

Note: If you plan to integrate with Claude Desktop, ensure your Claude config includes the bullet server entry as shown in the example.

Tips and caveats:

• Environment variables: You can control behavior with environment variables like BULLET_STRICT_MODE, BULLET_NO_CITATIONS, and BULLET_NO_COLOR as described in the repository documentation.
• For long documents, use the Sectioned Mode by providing a sections array with per-section items to get per-section scores.
• If citations are disabled, the output will omit research citations while preserving the scoring and feedback.
• Ensure your input context matches the intended usage (document, presentation, or reference) to optimize scoring and feedback behavior.
• If you encounter line-length or hierarchy warnings, consider restructuring lists to keep items concise and limit nested levels to two or fewer.