MCP server (rmcp stdio) exposing Discord tools to AI agents
28 tools with mostly complete schemas and descriptions, but significant gaps in parameter documentation and error handling guidance. Tool names follow verb_noun convention well (send, create, delete, list, read). Descriptions average ~80 chars, below the 194-char baseline for A+ tools. Most parameters lack detailed constraints (enums, ranges, formats). No output schemas documented. Error handling is minimal, no recovery guidance or actionable error messages. Security concerns: confirm flags are advisory only, not enforced server-side. Composition is reasonable (single responsibility per tool), but some tools like 'download' and 'top_reactions' operate on archived data with unclear dependencies.
Assign/remove a role to/from a member
List text/announcement/forum channels of a guild
Create a channel (admin)
Create a custom emoji (admin)
Create a role (admin)
Delete a channel (admin, gated)
Delete a custom emoji (admin, gated)
No output schemas documented. LLMs cannot predict response structure or plan downstream tool calls. E.g., 'read' returns messages but structure is undocumented; 'guilds' returns guild objects but fields are not specified.
Confirmation flags (confirm=true) are advisory only, not enforced server-side. Destructive operations (delete_channel, delete_role, delete_emoji, leave) can execute without actual confirmation, risking accidental data loss.
Parameter descriptions lack constraints. E.g., 'archive' accepts '60|1440|4320|10080' but no enum is defined; 'channel_type' accepts 'text|voice|category|announcement|stage|forum' but no enum constraint; 'status' in presence accepts 'online|idle|dnd|invisible' but no enum. LLMs will hallucinate invalid values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
Delete a role (admin, gated)
List DM + group-DM channels
Download archived attachments
Edit a channel (admin)
Edit a role (admin)
Get single message
List custom emojis
List joined guilds (name/id/icon/owner)
Join a server via invite
Leave a server
List roles
List members
Set presence (persisted; applies to next tail/watch connect)
Read recent messages (default 50) — the key AI-facing read
Search message
Send a message
Validate the configured token (exit 1 on failure)
Create a thread
List threads
Top-reacted messages from the archive
Show the authenticated user's profile
No error handling guidance. Tools return errors but do not tell LLMs what to do next (retry, ask user, abort). E.g., 'Invalid channel_id' does not suggest calling 'channels' to discover valid IDs.
Pagination not implemented. 'members' and 'search' accept 'limit' but no offset/cursor or total_count returned. Large result sets will blow context window; no way to iterate.