MCP server for generating and managing llms.txt documentation files
Server has 10 tools with mostly complete descriptions (194-char baseline met for primary tools). Tool naming follows verb_noun convention (generate_, validate_, update_, convert_, scan_, analyze_, health_, status_, help_, generate_). Schemas are present with typed parameters and descriptions. However, several tools lack output schema documentation, and some parameter descriptions are generic. Error handling guidance is minimal. Three utility tools (health_check, status, help) add noise without clear agent value. Missing tool annotations (readOnlyHint, destructiveHint, idempotentHint) for risk classification.
Analyze repository structure and metadata.
Convert llms.txt to LLM context format. This tool converts an llms.txt file into formats suitable for use as LLM context, supporting both XML and JSON output formats for integration with various LLM systems.
Generate llms.txt from a template.
Generate complete llms.txt documentation files for a project directory. This tool automatically scans a project directory, analyzes its structure and documentation files, and generates comprehensive llms.txt files that make the project AI-accessible. It intelligently categorizes documentation, source files, examples, and configuration files into structured sections following the llms.txt specification. The tool creates both a main llms.txt file (with links) and an llms-full.txt file (with full content) for maximum AI accessibility.
Perform a health check on the MCP server.
Get help information about available tools and their usage.
Missing output schema documentation for all tools. LLMs cannot plan downstream calls or extract required fields without knowing return structure (e.g., what fields does generate_llms_txt return? Is it a file path, content, or metadata object?).
Three utility tools (health_check, status, help) lack clear agent value and consume tool slots. health_check and status are server introspection; help duplicates MCP's built-in tool discovery. These should be removed or consolidated into a single diagnostic tool.
analyze_repo description is trivial ('Analyze repository structure and metadata.'), under 50 chars, fails to explain when to use it vs scan_project_structure. LLMs cannot disambiguate.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | <=2025-11-25 | v2 |
Scan and analyze project structure. This tool scans a project directory and returns detailed information about its structure, including file types, directory hierarchy, and project metadata.
Get the current status of the MCP server.
Update an existing llms.txt file. This tool updates an existing llms.txt file by regenerating specified sections while optionally preserving custom content. It allows for incremental updates without losing manually edited sections.
Validate an existing llms.txt file for format compliance and completeness. This tool performs comprehensive validation of an llms.txt file to ensure it follows the llms.txt specification and contains all required elements. It checks for proper structure, required sections, and data integrity.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. generate_llms_txt and update_llms_txt are WRITE operations; validate_llms_txt and scan_project_structure are READ_ONLY. Agents cannot infer safety without annotations.
Error handling descriptions missing. No guidance on retryability, user-fixable errors, or recovery steps. E.g., if generate_llms_txt fails due to unreadable directory, should the agent retry, ask the user, or try a different path?