A local markdown file management MCP server that allows AI agents to list, read, and write markdown files within a configured workspace directory, plus access current date/time information.
md-mcp has clear, action-verb naming (list_, read_, write_, get_) and all 4 tools have descriptions. However, descriptions are terse (avg ~60 chars vs. baseline 194), lacking context on WHEN to use each tool or dependencies. Input schemas are present and properly typed, but parameter descriptions are minimal. Output schemas are undocumented, LLMs cannot predict what fields to expect. Error handling is basic (no recovery guidance). Security is reasonable (path traversal checks in getSafePath), but no audit logging or permission gates.
Returns the current date and time in ISO 8601 format (UTC) and the local timezone.
Lists all the markdown files available in the workspace.
Reads and returns the complete text content of a specific markdown file.
Creates a new markdown file or overwrites an existing one with new content.
Output schemas undocumented. LLMs cannot predict return structure (fields, types, pagination). list_markdown_files returns a list but no schema describes the item structure or total count.
Descriptions are terse (avg 60 chars vs. baseline 194). Missing WHEN to use each tool and dependencies. E.g., read_markdown_file doesn't mention it requires a valid filename from list_markdown_files.
Error handling lacks recovery guidance. getSafePath returns 'security violation: access denied' but doesn't tell the LLM what to do next or suggest valid paths.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 66 | <=2025-11-25 | v2 |
No audit logging. Tool calls (especially write_markdown_file) are not logged with caller, timestamp, parameters, or outcome. Compliance and debugging are compromised.
write_markdown_file lacks confirmation/dry-run pattern. Destructive operations should support a confirmation step to prevent accidental overwrites.