Skip to content
Conduit

skill

discord-agent

Discord AI Agent bot with 45 tools, long-term memory, autonomous scheduler, and web capabilities. Powered by OpenRouter LLMs.

sitne/discord-agentcommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

The download comes from github.com/sitne/discord-agent itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 23, 2026

🤖 discord-agent

日本語

A Discord bot that acts as an autonomous AI agent with 79 tools and 5 skills, long-term memory, a cron-based task scheduler, web capabilities, code generation with GitHub CI, a vision system for idea-to-project tracking, and an extensible skills system following the SKILL.md standard. Powered by any LLM on OpenRouter.

~6,500 lines of Python across 17 files. Built with discord.py.

Features

  • LLM-powered agent — Routes through OpenRouter; switch models at runtime with /model
  • 79 tools across 11 categories (Discord, Thread, Forum, Memory, Scheduler, Web, System, CodeGen, HTTP, Skills, Vision)
  • Skills system — follows the SKILL.md standard with progressive disclosure; install community skills from GitHub, or let the bot create its own
  • Long-term memory — SQLite + FTS5 hybrid search with automatic context injection
  • Autonomous scheduler — Cron expressions, retry logic, dead-letter queue, execution history
  • Web access — Search, news, page reading (3-tier extraction + caching), and screenshots via Playwright
  • Conversation compression — Automatic summarization when context exceeds 20 messages
  • Safety first — Permission checks and confirmation buttons for destructive actions
  • Robust I/O — API retry with exponential backoff, parallel tool execution, image/text attachment processing
  • Vision system — Capture ideas, refine them into projects, track milestones, active projects auto-injected into context
  • MCP support — Connect external tool servers via mcp_servers.json

Tools

CategoryCountTools
Discord18send_message edit_message delete_messages create_channel delete_channel rename_channel set_channel_topic list_channels list_members get_member_info set_nickname create_role assign_role remove_role kick_member ban_member unban_member get_server_info
Thread7create_thread list_threads edit_thread delete_thread thread_add_member thread_remove_member send_thread_message
Forum5create_forum create_forum_post list_forum_posts manage_forum_tags edit_forum
Memory5remember recall forget forget_by_key list_memory_categories
Scheduler5create_scheduled_task list_scheduled_tasks delete_scheduled_task toggle_scheduled_task get_task_history
Web4web_search web_news read_webpage screenshot_webpage
System11run_shell + 10 GitHub CLI tools
CodeGen6codegen_create_project codegen_update_files codegen_check_ci codegen_list_projects codegen_read_file codegen_delete_file
HTTP1http_request
Skills6list_skills load_skill create_skill install_skill remove_skill search_community_skills
Vision8capture_idea list_ideas search_ideas update_idea create_project list_projects update_project project_dashboard

Setup

Requirements: Python 3.11+, uv

  1. Clone and install dependencies

    git clone https://github.com/sitne/discord-agent.git
    cd discord-agent
    uv sync
    uv run playwright install chromium
    
    python -m venv .venv && source .venv/bin/activate
    pip install -r requirements.txt
    playwright install chromium
    
  2. Configure environment

    cp .env.example .env
    

    Edit .env and set:

    DISCORD_TOKEN=your-bot-token
    OPENROUTER_API_KEY=your-api-key
    
    # Optional: override the default model
    # OPENROUTER_MODEL=openai/gpt-4o-mini
    
  3. Discord bot setup

    Create a bot at the Discord Developer Portal with the following privileged intents enabled:

    • Message Content
    • Server Members

    Invite the bot with permissions for managing channels, roles, messages, and members.

  4. Optional: Cloudflare & GitHub CLI setup (recommended)

    For web deployment via Cloudflare and GitHub integration:

    # GitHub CLI (repo management, CI, issues)
    # https://cli.github.com/
    gh auth login
    
    # Cloudflare CLI (Pages/Workers deploy)
    npm install -g wrangler
    # Set API token in .env (recommended for servers/bots)
    # Create at: https://dash.cloudflare.com/profile/api-tokens
    # → "Create Token" → "Edit Cloudflare Workers" template
    # Or for local dev: wrangler login (opens browser)
    
  5. Run

    uv run python bot.py
    

    Or with an activated venv: python bot.py

Commands

CommandDescription
/askSend a prompt to the agent (supports an optional image attachment)
/modelView or switch the active LLM (e.g. google/gemini-2.5-flash, openai/gpt-4o-mini)

📂 Project Structure

bot.py                  Entry point and bot initialization
db.py                   SQLite database, memory store, FTS5 search
tools.py                Discord management tools (18)
tools_web.py            Web search, reading, and screenshots
tools_system.py         Shell execution and GitHub CLI tools
tools_codegen.py        Code generation, GitHub project management, CI
tools_http.py           Generic HTTP client with SSRF protection
tools_skills.py         Skill management tools (SKILL.md standard)
tools_vision.py         Vision system: ideas & project tracking
tools_permissions.py    Permission checks and confirmation gates
skills_manager.py       Skill discovery, loading, install, and creation
skills/*/SKILL.md       Skill packages (5: cloudflare-deploy, data-analysis, image-gen, remotion-video, translation)
context_manager.py      Conversation history and compression
cron_parser.py          Cron expression parser
mcp_manager.py          MCP server lifecycle and tool routing
cogs/
  agent.py              Core agent loop, LLM calls, tool dispatch
  collector.py          Message and attachment collection
  scheduler.py          Cron scheduler, retry, DLQ, history

MCP Servers

To connect external Model Context Protocol tool servers, create a mcp_servers.json in the project root:

{
  "servers": [
    {
      "name": "example",
      "command": "npx",
      "args": ["-y", "@example/mcp-server"]
    }
  ]
}

MCP tools are automatically discovered and made available to the agent alongside built-in tools.

Dependencies

License

MIT

↑↓ navigate openesc close