MCP server for integrating with Raindrop.io bookmark management service
The server has 4 tools with consistent, clear naming (verb_noun pattern) and reasonable descriptions. However, critical deficiencies exist: (1) No input schemas are visible in the source code, only informal parameter lists in function signatures. The rubric requires explicit JSON Schema definitions with type declarations and descriptions. (2) Output schemas are completely undocumented, tools return formatted strings without specifying structure to downstream callers. (3) Error handling is minimal, no actionable recovery guidance for agents. (4) Parameter descriptions are present but lack formal constraints (enums, ranges, patterns). This is a mid-tier community server with good naming hygiene but incomplete schema documentation.
Add a new bookmark to Raindrop.io
Get latest bookmarks from Raindrop.io feed.
Search for bookmarks by keyword/text in Raindrop.io with optional date range filtering
Search for bookmarks with a specific tag in Raindrop.io with optional date range filtering
No explicit JSON Schema definitions visible in source code. Tools expose parameters as Python function args without formal type/constraint declarations. Rubric requires 'Every parameter needs a description explaining what it controls' AND formal schema with types. Input schemas appear inferred from function signatures rather than explicitly registered.
Output schemas completely undocumented. All tools return formatted strings via format_bookmark() helper. No specification of which fields are returned, their types, or their purpose. Downstream tools cannot reliably parse or chain results. Rubric requires 'Document the output schema. LLMs need to know what fields to expect.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | F | 43 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 53 | - | v1 |
Error handling provides no recovery guidance. Generic messages like 'Unable to fetch bookmarks or no bookmarks found' give agents no next steps. Rubric requires 'Error responses must tell the LLM what to do next' and categorize errors as retryable/user-fixable/fatal. No distinction between API errors, auth failures, and no-results conditions.
No enum constraints or range validation declared for parameters. The 'count' parameter accepts any integer without declared bounds (rubric: 'page_size 1 - 100'). 'collection_id' has no constraints. 'from_date'/'to_date' accept free-form strings without regex pattern or validation hint beyond 'YYYY-MM-DD format'. Descriptions mention format but no formal constraint is enforced or declared.
Parameter descriptions are present but brief (mostly 10-50 chars) and lack critical context. 'count' says 'Number of bookmarks to fetch (default: 10)' but not the min/max bounds or what happens if exceeded. 'collection_id: ID of the collection' doesn't explain where to find valid IDs or what '0' means in Raindrop API terms. Rubric: descriptions should be 50-200 chars and state 'expected format, range, and allowed values.'
RAINDROP_TOKEN stored in environment variable and used globally. No explicit secret injection pattern declared. While not exposed as a tool parameter (correct), the implementation does not document the secret management story. Rubric requires 'Credentials must never appear as tool parameters. Use server-side secret injection via environment variables or vault.' This is acceptable but should be explicit in docs.