Discord MCP Server
Enables AI models to interact with Discord using human-readable channel names and usernames through smart target resolution and automatic mention processing. It provides tools for sending messages, reading channel history, and searching for content without requiring manual snowflake ID lookups.
README
Discord MCP Server
A streamlined Model Context Protocol server for Discord with smart target resolution. No more hallucinated IDs!
Features
- 🎯 Smart Target Resolution: Use channel names, usernames, or IDs - no need to memorize snowflakes
- 🧠 LLM-Friendly: Reduces hallucination by accepting human-readable names
- 💬 Automatic Mention Processing: Converts @username to proper Discord
<@id>format automatically - 📦 Streamlined API: Only 7 essential tools, no redundancy
- 💾 Smart Caching: Automatically caches name→ID mappings
- 🔀 Ambiguity Handling: Detects when channel names collide and guides to use
ServerName/channelformat
Tools
- send_message - Send to channels or DMs (accepts names or IDs)
- edit_message - Edit or delete messages (empty content = delete)
- read_messages - Read channel history + channel info
- list_servers - List all accessible servers
- list_channels - List channels in a server
- search_messages - Search for messages in a channel
- add_reaction - React to messages with emoji
Setup
1. Create a Discord Bot
- Go to Discord Developer Portal
- Click "New Application" and give it a name
- Go to "Bot" section and click "Add Bot"
- Enable these Privileged Gateway Intents:
- Message Content Intent
- Server Members Intent
- Presence Intent (optional)
- Click "Reset Token" and copy your bot token
- Go to "OAuth2" → "URL Generator"
- Select scopes:
bot - Select permissions:
Send Messages,Read Message History,Add Reactions,Manage Messages
- Select scopes:
- Use the generated URL to invite the bot to your server
2. Install Dependencies
pip install -r requirements.txt
3. Set Environment Variable
export DISCORD_TOKEN="your_bot_token_here"
Or on Windows:
set DISCORD_TOKEN=your_bot_token_here
4. Run the Server
python path/to/discord-mcp
Usage Examples
With Claude Desktop (config)
Add to your claude_desktop_config.json:
{
"mcpServers": {
"discord": {
"command": "python",
"args": ["/path/to/discord-mcp"],
"env": {
"DISCORD_TOKEN": "your_bot_token_here"
}
}
}
}
Example Prompts for Claude
Send a message:
Send "Hello everyone!" to the general channel
Send "Meeting in 5 mins" to Work Server/announcements
Read messages:
Read the last 20 messages from announcements
Read messages from Gaming Server/general
Search:
Search for messages containing "meeting" in the team-chat channel
Search for "bug report" in Dev Team/bugs
Edit/Delete:
Edit message 123456789 to say "Updated: Meeting at 3pm"
Delete message 987654321
Smart Target Resolution
The server automatically handles both names and IDs:
| Input Type | Example | How It Works |
|---|---|---|
| Channel name | "general" |
Searches for channel by name |
| Server/Channel | "MyServer/general" |
Searches in specific server (solves ambiguity) |
| Channel with # | "#announcements" |
Strips # and searches by name |
| Username | "john" |
Searches for user by username |
| Username with @ | "@alice" |
Strips @ and searches by username |
| Snowflake ID | "123456789012345678" |
Uses ID directly (17-20 digits) |
Handling Ambiguous Channel Names
Since most Discord servers have channels with common names like "general" or "announcements", the server handles ambiguity intelligently:
If a channel name is unique: Just use the name
Send "Hello!" to announcements
If a channel name appears in multiple servers: The server will tell you which servers have that channel and ask you to specify:
Error: Multiple channels named 'general' found:
• My Gaming Server → #general
• Work Team → #general
• Friend Group → #general
Please specify format: 'ServerName/channel' or use channel ID
Use the ServerName/channel format:
Send "Hello team!" to Work Team/general
Read the last 10 messages from My Gaming Server/general
This completely eliminates the need for the AI to remember or hallucinate long channel IDs!
Automatic User Mention Processing
Discord bots can only mention users using the <@user_id> format, but LLMs naturally want to use @username. The server automatically handles bidirectional conversion:
When SENDING messages (AI → Discord):
AI writes: "Hey @john, can you check this?"
Discord receives: "Hey <@789012345678901234>, can you check this?"
When READING messages (Discord → AI):
Discord has: "Meeting with <@789012345678901234> at 3pm"
AI sees: "Meeting with @john at 3pm"
Sending - Handles:
@username→ Looks up user and converts to<@id>@123456789→ Recognizes as ID and formats to<@123456789>123456789→ Detects raw IDs and converts to<@123456789>if valid user- Non-existent users → Left as plain text (won't create broken mentions)
Reading - Handles:
<@123456789>→ Fetches user and converts to@username<@!123456789>→ Handles nickname format, converts to@username- Unknown user IDs → Shows as
@[123456789](fallback format)
This bidirectional conversion means:
- ✅ The AI can write natural messages with @mentions
- ✅ The AI can read and understand who's being mentioned in chat history
- ✅ The AI can quote or reference previous mentions correctly
- ✅ No confusion with long user IDs
Caching
The server caches name→ID mappings to improve performance and reduce API calls. Cache is maintained in memory during runtime.
Architecture
discord_mcp_server.py
│
├─ parse_target() - Parses "ServerName/channel" format
├─ process_mentions() - Converts @username and raw IDs to <@id> format (AI → Discord)
├─ humanize_mentions() - Converts <@id> back to @username format (Discord → AI)
├─ standardize_server() - Resolves server names/IDs to Guild objects
├─ standardize_channel() - Resolves channel names/IDs to Channel objects
│ Returns (channel, error) for ambiguity handling
├─ standardize_user() - Resolves usernames/IDs to User objects
│
└─ MCP Tools:
├─ send_message - Uses parse_target(), process_mentions(), standardize functions
├─ edit_message - Uses process_mentions() + direct message ID lookup
├─ read_messages - Uses parse_target(), humanize_mentions(), standardize functions
├─ list_servers - No resolution needed
├─ list_channels - Uses standardize_server()
├─ search_messages - Uses parse_target(), humanize_mentions(), standardize functions
└─ add_reaction - Direct message ID lookup
Error Handling
The server provides clear error messages:
- "Could not find channel 'xyz'" - Channel name/ID not found
- "Could not find server 'xyz'" - Server name/ID not found
- "Could not find message with ID xyz" - Message doesn't exist or bot lacks access
Permissions
Ensure your bot has these permissions:
- Read Messages/View Channels
- Send Messages
- Read Message History
- Add Reactions
- Manage Messages (for editing/deleting)
Troubleshooting
Bot not responding:
- Check that
DISCORD_TOKENis set correctly - Verify bot is invited to the server
- Ensure bot has necessary permissions
"Could not find channel" errors:
- Check channel name spelling
- Verify bot has access to the channel
- Try using the channel ID instead
Message edit/delete fails:
- Bot can only edit/delete its own messages
- Ensure the message ID is correct
- Check bot has "Manage Messages" permission
License
APACHE LICENSE 2.0 - feel free to modify and use as needed!
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。