raw-discord-mcp

Retrieves raw message data from Discord channels and servers, converting Discord API responses into structured context that can be injected into MCP-compatible LLM clients. Uses Discord.py or similar library bindings to authenticate via bot token and fetch message history, metadata, and thread information, then normalizes the output into MCP resource format for seamless integration with Claude, other LLMs, or AI agents.

Discovers and catalogs all accessible Discord channels, servers, and threads within a bot's permission scope, extracting metadata including channel type (text, voice, thread), topic, permissions, member count, and creation timestamps. Implements recursive traversal of guild hierarchies and category structures, normalizing channel relationships into a queryable resource tree that MCP clients can browse and reference.

Implements server-side message search across Discord channels using Discord's native search API or local indexing, supporting filters by author, date range, content keywords, and message type. Returns paginated result sets with relevance scoring, enabling LLMs to locate specific conversations or threads without loading entire channel histories. Handles Discord's search syntax (from:, before:, after:, has:) and translates them into MCP query parameters.

Enables MCP clients to send, edit, and delete messages in Discord channels through the MCP protocol, with support for embeds, attachments, and formatting. Implements Discord's message object schema (content, embeds, components) and handles rate limiting, permission validation, and error recovery. Allows LLMs to take actions in Discord (respond to conversations, post updates, correct mistakes) as part of agentic workflows.

Fetches user profiles, member information, and role assignments from Discord servers, including username, avatar, status, roles, join date, and custom metadata. Implements caching to reduce API calls and supports bulk member queries for large servers. Normalizes user data into a consistent schema that LLMs can reference when analyzing conversations or making decisions based on user context.

Manages message reactions and emoji interactions, allowing LLMs to add reactions to messages, retrieve reaction counts, and interpret emoji reactions as user feedback or voting signals. Supports both standard Unicode emojis and custom Discord server emojis, with caching of emoji metadata. Enables feedback loops where LLM responses can be validated through Discord reactions.

Manages Discord threads (both public and private), enabling creation, retrieval, and message posting within threaded conversations. Implements thread metadata extraction (creator, creation time, archive status) and supports archiving/unarchiving threads. Allows LLMs to organize conversations into threads and maintain separate context for parallel discussions within a single channel.

Implements webhook-based event streaming from Discord to the MCP server, enabling real-time notification of messages, reactions, member joins, and other server events. Uses Discord's gateway connection or webhook payloads to push events to the MCP server, which can then trigger LLM actions or update context. Handles event deduplication and reconnection logic for reliable event delivery.

Validates bot and user permissions before executing actions, checking channel permissions, role-based access, and message author authorization. Implements Discord's permission bitfield system and role hierarchy, preventing unauthorized actions and providing clear error messages when permissions are insufficient. Enables LLMs to make permission-aware decisions about which actions are feasible.