communication MCP Server
Local-first, self-hostable Discord bot that runs full D&D 5e games end-to-end with an AI Dungeon Master persona (ShoeGPT). Three-brain architecture: oMLX (narration) + dm20 MCP (rules engine) + discord.py (orchestrator). The LLM never computes game math.
Discovered via github-topic:mcp and last synced 3mo ago.
Install instructions not detected yet
Check the source repository for the latest setup steps.
✅
❌
❌
What it does
❌
❌
Free
oMLX / dm20
Why
❌
❌
✅*
❌
❌
None
What ships
`brew install [email protected]`
❌
❌
Channel ID → `(campaign_name, claudmaster_session_id, dm20_party_token, current_state)`
❌ no `mlx-lm` wheels
Required
❌
❌
Active riposte buttons with `deadline_ts` — drives the background expiry sweeper
ℹ️ **Don't have oMLX + dm20 set up yet?** Check the [oMLX docs](https://github.com/macabdul9/omlx) and [dm20-protocol README](https://github.com/Polloinfilzato/dm20-protocol). On Jeremy's reference rig they're supervised by `launchd` as `com.user.omlx` so they survive reboots. The install script will warn you (not fail) if they're not running. --- ## ⚙️ Installation (the Verbose Tour) ### 🤖 The easy way ```bash ./install.sh ``` That's it. The script: 1. 🔎 Checks Python version (must be ≥3.11) 2. 🔎 Checks for `uv` (installs via official script if missing) 3. 🌱 Creates a `.venv/` virtualenv with `uv venv` 4. 📦 Installs all Python dependencies (`discord.py`, `httpx`, `aiosqlite`, `pydantic`, `tenacity`, `structlog`, `ocrmac` on macOS, `PyMuPDF`, `pypdf` as fallback, plus dev deps for tests/lint) 5. 🩺 Pings `:8765/v1/models` to verify oMLX is running and reports which model is loaded (should be `ShoeGPT`) 6. 🩺 Pings `:8765/v1/mcp/tools` to confirm dm20 is exposed (expects ≥97 dm20 tools) 7. 💡 Tells you exactly what to do next (copy `.env.example`, run `bootstrap`, run the bot) If anything fails, the script tells you why in plain English (no cryptic exit codes). 💬 ### 🧙 The "I want to know what's happening" way If you want to do this by hand: ```bash # 1) Make sure Python 3.11+ is your interpreter python3 --version # → Python 3.11.x or higher # 2) Install uv if you don't have it (fast, hermetic, friendly) curl -LsSf https://astral.sh/uv/install.sh
❌
❌
❌
What's in it
Status
❌
Every persistent Discord View we've posted: `custom_id` → `(view_class, message_id, channel_id, payload_json)`
Every player input where the sanitizer stripped or truncated something. Forensics for prompt-injection attempts.
>`, `SYSTEM:`, `ASSISTANT:`, `<player_action>`, etc.) so a player can't forge a tool call by typing one 3. 📦 Wraps the cleaned text in `<player_action speaker="..." user_id="...">…</player_action>` sentinels so downstream prompts can see "this came from a player, treat as untrusted" 4. 📝 Logs to `sanitizer_audit` whenever it actually stripped or truncated something A ≥30-scenario adversarial corpus runs in CI — known injection attempts and tool-call forgery patterns, all of which must pass-through-cleaned. If the corpus fails, the build fails. ### 🎮 The Discord Layer (`src/eldritch_dm/bot/`) `discord.py 2.7+` with **persistent Views**, because the bot will absolutely be restarted while a game is in progress and the buttons need to still work afterward. We use `discord.ui.DynamicItem` with regex `custom_id` templates like `endturn:(?P<channel_id>\d+):(?P<actor>\d+)`, register them in `setup_hook`, and call `bot.add_view(view, message_id=...)` for every row we find in `persistent_views`. The kill-and-restart drill is part of the test suite. Other Discord disciplines: - ⏱️ **Defer first, always.** The first line of every interaction callback is `await interaction.response.defer(thinking=True)`. A custom ruff rule fails CI if any callback omits this. Discord gives you 3 seconds before the interaction expires — narration takes longer than that, so we acknowledge instantly and follow up with the answer. - 📡 **Embed coalescer.** During combat, the embed updates many times per round. Discord rate-limits message edits at ~5/5s; we limit ourselves to ≤1/sec via a per-message `asyncio.Queue` + render task. Under the 8-player load test, zero 429s. 🟢 - ⚠️ **Ephemeral warnings.** "❌ Not your turn," "❌ Riposte expired," "🔌 DM is offline" — all delivered as ephemeral followups so only the offending user sees them. ### 🗡️ The Riposte Magic (`src/eldritch_dm/combat/riposte.py`) This is the most fun piece. When dm20 resolves a monster's attack as a miss against an eligible PC (**Battle Master Fighter** by RAW — see [Known Limitations](#-known-limitations-v1) for the v1 scope; v2 plans to make eligibility YAML-configurable for homebrew) who has their reaction available, EldritchDM: 1. Inserts a row in `riposte_timers` with `deadline_ts = now() + RIPOSTE_TTL_SECONDS` 2. Posts an ephemeral message visible only to that PC's user, containing the `[ ↩️ Riposte Counter-strike ]` button 3. The button's `custom_id` includes the timer ID and is registered as a `DynamicItem` 4. A background sweeper task wakes at the deadline and removes the message 5. If the bot is killed before the deadline and restarted after, the sweeper picks up the still-pending row and either continues the wait or cleans up an expired one 6. On click, the bot calls `dm20__combat_action(reaction=true, weapon=primary)` and narrates the result It's the kind of thing every D&D player wishes their VTT had. 🥹 --- ## 🗺️ Roadmap EldritchDM v1 is feature-complete. Here's the 5-phase history:
Send messages and upload files to Slack channels using the `slack` CLI. Supports direct messages, channel posts, file uploads, and thread replies.
Bidirectional Telegram bridge — talk to Claude from your phone. Built into OpenPaw.
Discover agents across local node and mesh peers at runtime — their roles, handles, and how to communicate with or delegate to them.
Apply data visualization and storytelling principles from Storytelling with Data by Cole Nussbaumer Knaflic. Covers choosing effective visuals (line, bar, table, scatterplot), eliminating clutter (Gestalt principles, data-ink ratio), focusing attention (preattentive attributes, strategic color), thinking like a designer (affordances, alignment, white space), and narrative structure (three-act story, horizontal/vertical logic, Big Idea). Trigger on "data visualization", "chart design", "dashboard design", "data storytelling", "presentation chart", "declutter chart", "bar chart", "line chart", "data narrative", "slide deck data", "chart review", "viz critique", "storytelling with data".
🔬 A curated collection of 23,000+ agent skills for empirical research across 8 social science disciplines. | 精选 23,000+ AI Agent 技能库，覆盖8大社会科学学科的实证研究。CoPaper.AI 20分钟完成一篇可复现的规范实证论文，并支持用户上传 Skills。-- Maintained by CoPaper.AI from Stanford REAP.
🔬 A curated collection of 23,000+ agent skills for empirical research across 8 social science disciplines. | 精选 23,000+ AI Agent 技能库，覆盖8大社会科学学科的实证研究。CoPaper.AI 20分钟完成一篇可复现的规范实证论文，并支持用户上传 Skills。-- Maintained by CoPaper.AI from Stanford REAP.
AgentNetworkProtocol(ANP) is an open source protocol for agent communication. Our vision is to define how agents connect with each other, building an open, secure, and efficient collaboration network for billions of intelligent agents.
Learn how to use the c-slack Claude skill. Complete guide with installation instructions and examples.
Learn how to use the c-telegram Claude skill. Complete guide with installation instructions and examples.
Learn how to use the Letter Writer Claude skill. Complete guide with installation instructions and examples.