Discord Selfbot

quick start

this is a local mcp server (stdio transport). your mcp client spawns it as a process, and you provide secrets (like DISCORD_TOKEN) via the client's env/environment config.

manual run:

DISCORD_TOKEN="your_discord_token_here" python3 -m discord_py_self_mcp.main

important: automating user accounts is against the Discord ToS. use this at your own risk.

overview

discord-py-self-mcp acts as a bridge between your ai assistant (Claude Code, OpenCode, Codex, etc) and your personal discord account. unlike standard bots, this "selfbot" runs as you; allowing your ai to read your dms, reply to friends, manage your servers, and interact with buttons/menus just like a human user.

built on the discord.py-self library by dolfies.

quick installation

paste this into your llm agent session:

Install and configure discord-selfbot-mcp by following the instructions here:
https://raw.githubusercontent.com/Microck/discord.py-self-mcp/refs/heads/master/INSTALL.md

npm (recommended)

npm install -g discord-selfbot-mcp
discord-selfbot-mcp-setup

manual installation

prerequisites:

• python 3.10+
• uv (recommended) or pip
• voice support (linux only): libffi-dev (or libffi-devel), python-dev (e.g. python3-dev)

install:

uv tool install git+https://github.com/Microck/discord.py-self-mcp.git
# or
pip install git+https://github.com/Microck/discord.py-self-mcp.git

note: voice dependencies (PyNaCl) are included by default. on linux, ensure system packages are installed first.

npm installation (node.js wrapper)

prerequisites:

• node.js 18+
• python 3.10+

install:

npm install -g discord-selfbot-mcp

the npm package is a wrapper that uses the underlying python implementation.

how it works (setup wizard)

run the interactive setup script to extract your token and generate the mcp config json (it can also write to common client config files and creates a backup before editing).

# if using npm
discord-selfbot-mcp-setup

# if using python (uv/pip)
python3 -m discord_py_self_mcp.setup

1. extract token: grabs your token from an open browser session (playwright) or via manual entry
2. generate config: prints the mcp configuration json (and can write it to your client config)
3. configure: paste the config into your mcp client settings

manual configuration

because this server uses stdio, you configure it as a local command and pass the token via env (not url/headers).

examples:

• mcp.example.json
• mcp.python.example.json
• .env.example

npm wrapper (recommended):

{
  "mcpServers": {
    "discord-py-self": {
      "command": "discord-selfbot-mcp",
      "env": {
        "DISCORD_TOKEN": "${DISCORD_TOKEN}"
      }
    }
  }
}

python (uv tool):

{
  "mcpServers": {
    "discord-py-self": {
      "command": "uv",
      "args": ["tool", "run", "discord-py-self-mcp"],
      "env": {
        "DISCORD_TOKEN": "${DISCORD_TOKEN}"
      }
    }
  }
}

python (pip / venv):

{
  "mcpServers": {
    "discord-py-self": {
      "command": "python3",
      "args": ["-m", "discord_py_self_mcp.main"],
      "env": {
        "DISCORD_TOKEN": "${DISCORD_TOKEN}"
      }
    }
  }
}

if your client does not expand ${DISCORD_TOKEN}, replace it with the literal token value.

features

powered by the robust discord.py-self library.

discrawl integration

Use run_discrawl to execute local discrawl commands directly from MCP. Use typed tools for common operations (discrawl_sync, discrawl_search, discrawl_messages, etc.) when you want structured params. This fork prefers a sibling checkout at ../discrawl-self/bin/discrawl when present, then falls back to discrawl in PATH.

Example tool call payload:

{
  "command": "sync",
  "args": ["--guild", "1234567890", "--since", "2026-03-01T00:00:00Z"],
  "config_path": "~/.discrawl/config.toml"
}

Typed tool payload example:

{
  "tool": "discrawl_sync",
  "args": {
    "guild": "1234567890",
    "since": "2026-03-01T00:00:00Z",
    "full": true,
    "config_path": "~/.discrawl/config.toml"
  }
}

Optional env var:

• DISCRAWL_BIN - custom path to discrawl executable. This overrides the default fork-first lookup.

attachment access

Use get_message_attachments when a message contains files or images you need to inspect directly. It returns attachment metadata for the target message and can stream image/file content back through MCP outputs.

comparison

✅ = supported

❌ = not supported

🚧 = planned / in progress

captcha solving (experimental)

automatically solves hCaptchas when encountered (e.g., joining servers, dms). built upon QIN2DIM/hcaptcha-challenger - an AI-powered hCaptcha solver using vision models.

warning: this feature is experimental. use at your own risk.

requirements:

1. Gemini API Key: Required for AI vision. Get from Google AI Studio. Set GEMINI_API_KEY in your mcp client env.
2. Playwright: Required for browser automation.

   playwright install chromium --with-deps
   

optional:

• CAPTCHA_PROXY: proxy url for solving hCaptcha challenges.
• TEMP_DIR: Directory for temporary model files (default: /tmp/hcaptcha)

rate limiting (recommended)

built-in rate limiting to prevent account bans. configurable via environment variables.

configuration:

rate limiting is enabled by default to reduce ban risk. Only disable it if you are deliberately taking responsibility for raw Discord API pacing yourself.

troubleshooting

project structure

discord_py_self_mcp/
├── bot.py
├── main.py
├── setup.py
├── rate_limiter.py
├── tool_utils.py
├── cli_runtime.py
├── logging_utils.py
├── captcha/
│   └── solver.py
└── tools/
    ├── channels.py
    ├── discrawl.py
    ├── embed.py
    ├── guilds.py
    ├── interactions.py
    ├── invites.py
    ├── members.py
    ├── messages.py
    ├── presence.py
    ├── profile.py
    ├── reactions.py
    ├── registry.py
    ├── relationships.py
    ├── threads.py
    └── voice.py

skill cli mode (optional)

in addition to the mcp server, this package also provides a skill/cli mode for command-line usage with a persistent daemon. this is useful for scripts or when you need faster execution without the mcp protocol overhead.

quick start:

# install (same package)
npm install -g discord-selfbot-mcp

# create .env file
echo "DISCORD_TOKEN=***" > .env

# use skill mode (from package directory)
python3 scripts/dcli.py send-message --channel 123 --content "Hello!"

key commands:

python3 scripts/dcli.py daemon start     # start the daemon
python3 scripts/dcli.py daemon status    # check daemon status
python3 scripts/dcli.py send-message --channel CHANNEL_ID --content "Hello"
python3 scripts/dcli.py list-guilds
python3 scripts/dcli.py read-messages --channel CHANNEL_ID --limit 20
python3 scripts/dcli.py get-message-attachments --channel CHANNEL_ID --message MESSAGE_ID
python3 scripts/dcli.py get-message-attachments --channel CHANNEL_ID --message MESSAGE_ID --download --output-dir ./attachments

when to use skill mode:

• command-line scripting
• faster execution (persistent connection)
• automation tasks
• when mcp is not needed

see SKILL.md for detailed documentation.

license

this project is licensed under the mit license.

contributing

issues and pull requests are welcome at github.com/Microck/discord.py-self-mcp.

1. fork the repository
2. create a feature branch (git checkout -b feature/my-feature)
3. commit your changes (git commit -m 'add my feature')
4. push to the branch (git push origin feature/my-feature)
5. open a pull request

please ensure tests pass (pytest) before submitting.