This server provides tools for downloading works from RedNote. A URL containing any of the following strings is a RedNote work URL: 1. www.xiaohongshu.com 2. xhslink.com Call the download_rednote_work() function to download the work via its URL. Call the get_rednote_work_info() function to get the work info via its URL.
Two tools with basic structure but significant quality gaps. Naming is action-verb compliant (download_, get_), but descriptions lack LLM-optimization guidance. Input schemas are present but minimal (single string parameter). Output schemas are completely undocumented, both tools return plain strings rather than structured objects. Error handling is absent: the tool returns translated error messages as strings with no guidance for agent recovery. The server relies on external domain knowledge (RedNote URL patterns) documented in instructions, not in parameter descriptions. No pagination, no parameter constraints (URL format), no idempotent hints, no rate limiting advice.
Download works from RedNote using the work's URL. The works include pictures, videos, and LivePhotos.
Retrieve RedNote post information, including: - Favorites count - Comments count - Shares count - Likes count - Work tags - Work ID - Work URL - Work title - Work description - Work type - Publication time - Last update time - Timestamp - Author nickname - Author ID - Author URL - Download link - Animated image URL - And other relevant details It is crucial to note the following: The Download link in the data returned by this MCP tool cannot be used as the "url" input for the download_rednote_work tool. If you intend to use the download_rednote_work tool to download the work, you must use the original URL entered by the user.
Output schema completely undocumented. Both tools return plain strings instead of structured objects with typed fields. LLM cannot extract data for downstream tool chaining or parsing.
No parameter validation or constraints. 'url' parameter lacks format specification, regex pattern, or examples. Description says 'The URL of the work' but does not state accepted formats (www.xiaohongshu.com, xhslink.com), that critical info is buried in server instructions, not in parameter docs.
Error handling returns translated error messages as plain text strings. No recovery guidance for agent: 'Download failed' tells the LLM nothing about why it failed or what to try next. No categorization of errors as retryable vs user-fixable vs fatal.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 47 | - | v1 |
get_rednote_work_info description is 380+ characters and includes prescriptive constraints ('The Download link in the data returned... cannot be used'). This confuses the agent by stating what NOT to do rather than describing what the tool outputs. Should be 50 - 200 chars and let the output schema specify field types.
Both tools accept only a single 'url' parameter. No pagination support, no way to batch-download multiple works, no limit parameter. If an agent needs to fetch 10 post details, it must make 10 sequential calls.
Tool annotations missing. download_rednote_work has destructiveHint=true (downloads/writes files) and idempotent=false (repeated calls with same URL trigger re-downloads), but no hints in the registration. get_rednote_work_info is read-only, should declare readOnlyHint=true for agent planning.