All work
AI Agents & MCP

Discord MCP

An MCP server that lets Claude manage a Discord server and catch scams transparently.

Channel management plus anti-scam moderation with explainable scoring and a dry-run-then-confirm purge workflow.

What makes it serious
  • Channel management driven from Claude
  • Explainable scam-scoring heuristics, not a black box
  • Dry-run scan then confirmed-purge workflow
  • Human-in-the-loop before any removal

A Python MCP server for interacting with and moderating a Discord server from Claude. Alongside channel management, its standout feature is transparent anti-scam moderation: it scores members and messages on explainable heuristics — account age, phishing keyword-plus-link combinations, and cross-channel blast patterns — rather than an opaque model. Nothing is destructive by default; a dry-run scan surfaces suspects for review, and only a confirmed step carries out a purge.

The transparency is deliberate. Moderation decisions affect real people, so the reasoning behind each flag is legible and the human stays in the loop before anyone is removed. It is built on discord.py and the MCP protocol. It is my one confirmed-public MCP on GitHub, and the moderation heuristics are written to be tuned and audited rather than trusted blindly, so a server owner can see precisely why a member was flagged before acting.

Stack
Pythondiscord.pyMCP
#mcp#discord#moderation#anti-scam#claude#python