A Model Context Protocol server for fetching ietf documents for LLMs
This server has three tools with explicit MCP registration via @mcp.tool() decorators. However, critical gaps in parameter descriptions, missing output schemas, and limited error handling significantly reduce quality. Tool naming is functional but not optimally specific. No input validation or error guidance. The server is STDIO-only, which is a separate transport limitation (hard-capped at 50 for protocol readiness).
Get an RFC document by its number in RFC editor Index with pagination support
Get the total number of IETF RFC documents available in RFC editor Index
Search for IETF RFC documents from RFC Editor Index by keyword in their titles
Output schemas are completely undocumented. list_ietf_docs_number returns an int, but get_ietf_doc returns a Dict[str, Any] with no field documentation. search_ietf_rfc_by_keyword returns List[Dict[str, str]] but the agent does not know what fields 'number' and 'title' contain or their types. Without documented output schemas, LLMs cannot plan downstream tool calls or extract the correct data.
get_ietf_doc accepts parameters 'start_line' and 'max_lines' for pagination, but there is no documented total count or next_cursor in the response. An LLM cannot determine whether more results exist, causing incomplete pagination loops. Tool should return {'content': '...', 'start_line': int, 'max_lines': int, 'total_lines': int, 'has_more': bool} to enable proper iteration.
No error handling or recovery guidance. If an RFC number does not exist, the agent receives a failure with no actionable next step. Tool descriptions do not explain what to do if a search returns zero results, if a download fails, or if the cache is corrupted. LLMs need explicit recovery instructions.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 49 | <=2025-11-25 | v2 |
| 2026-03-09 | D | 53 | - | v1 |
Parameter 'number' in get_ietf_doc is described as 'The RFC number str (e.g., "1234")' but declared as type integer. The description contradicts the schema. Additionally, the description uses an example value ('1234') which LLMs may treat literally. Use constraints instead: 'RFC number (1 - 9999)' with type integer.
Parameter descriptions lack actionable constraints. 'max_lines' has no stated range (is 1 the minimum? Is there a maximum?). 'keyword' in search_ietf_rfc_by_keyword has no length limit or case-sensitivity guidance. LLMs will pass unbounded or invalid values.
Tool naming could be more specific. 'list_ietf_docs_number' is awkward phrasing (noun 'number' follows verb 'list'); clearer as 'count_ietf_rfc_documents' or 'get_rfc_count'. 'get_ietf_doc' is ambiguous, does it get metadata, full text, or headers? Rename to 'get_rfc_document_text' or 'fetch_rfc_content'.
No input validation. Tool implementations accept raw user input (RFC number, keyword) without bounds checking. A malformed RFC number like '999999999' or a massive keyword string will be passed to downstream functions without validation, risking crashes or resource exhaustion.
Tool descriptions do not mention prerequisites or when to use one tool vs another. For example, should the agent call list_ietf_docs_number first to validate that an RFC exists before calling get_ietf_doc? This guidance is missing.