HTML entities MCP. Keyless, offline: encode text to HTML entities and decode HTML entities back to text (named + numeric decimal/hex). Pure functions — no API, no key.
Two simple, well-named tools with clear descriptions and complete input schemas. Both tools are READ_ONLY, offline, and keyless. Naming follows verb_noun convention (encode_html, decode_html). Descriptions are concise (90 - 130 chars) and explain what each tool does. Input schemas include type and description for all parameters. However, output schemas are not documented, LLMs cannot predict the structure of returned objects. Error handling is minimal (only 'Unknown tool' error visible). No parameter validation guidance or enum constraints for the 'mode' parameter in encode_html.
Decode HTML entities (named like & and numeric like © / ©) back to plain text. Keyless, offline.
Encode text to HTML entities. mode "basic" (default) escapes only & < > " '; mode "extended" also converts every non-ASCII character to a numeric entity. Keyless, offline.
Output schemas not documented. LLMs cannot predict return structure (e.g., encode_html returns {input_length, mode, encoded}; decode_html returns {decoded}). Without documented output schemas, agents cannot plan downstream operations or extract fields reliably.
encode_html 'mode' parameter lacks enum constraint. Description says '"basic" (default) or "extended"' but no enum is declared in inputSchema. LLMs may hallucinate invalid values like 'minimal' or 'full'. Should add enum: ['basic', 'extended'].
Minimal error handling. Only 'Unknown tool' error is visible. No validation of input types, no guidance on what to do if 'text' is missing or malformed. Error messages should be actionable (e.g., 'Required argument "text" is missing (a string).' is good, but should be returned as a structured error response, not thrown).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
No documentation of parameter constraints. 'text' parameter has no length limits, encoding hints, or examples. For encode_html, should clarify: 'The text to encode (any length; non-ASCII chars handled per mode).' For decode_html: 'Text containing HTML entities (named like &, numeric like © or ©).'