communication MCP Server
🦅 Open-source AI agent runtime in Rust — Self-hostable with persistent memory, self-improving skills, MCP support, built-in cron, multi-platform messaging & production gateway.
Discovered via github-topic:mcp and last synced 1w ago.
Install instructions not detected yet
Check the source repository for the latest setup steps.
Required tokens
Same as `smart` — for trusted automation pipelines
`SLACK_BOT_TOKEN`, `SLACK_APP_TOKEN`
`/v1/responses` endpoint
Control Chrome/Chromium via CDP — click, type, screenshot, run JS
Spawn a parallel sub-agent for decomposed work (depth-limited by `max_delegation_depth`)
`FileMemoryStore` (markdown) + `SessionDb` (SQLite + FTS5)
Example
`TELEGRAM_TOKEN`
`anthropic`
Search the web (Brave / Serper / DuckDuckGo)
Read and update the persistent user profile
LLM adapters: Anthropic, OpenAI-compat, Bedrock, Codex, Ollama, vLLM, ThaiLLM
`garudust-*-x86_64-pc-windows-msvc.zip`
Blocks all destructive calls — for read-only agents
`MATRIX_HOMESERVER`, `MATRIX_USER`, `MATRIX_PASSWORD`
`ollama` *(add `url:` for custom endpoint)*
Read a file from the filesystem
List available skills
Agent run loop, context compressor, prompt builder
ollama
`providers.default.name`
Fetch a URL
Persistent key-value memory across sessions
Shared traits & types — zero I/O
Send message
output format, language, tone, tool choices
`WHATSAPP_ACCESS_TOKEN`, `WHATSAPP_PHONE_NUMBER_ID`, `WHATSAPP_VERIFY_TOKEN`
`thaillm`
List files with glob patterns and depth limits
Create or update a skill in `~/.garudust/skills/`
Cron scheduler
Action
Behaviour
--- ## Contents - [Why Garudust?](#why-garudust) - [Supported Platforms](#supported-platforms) - [Quick Start](#quick-start) - [Architecture](#architecture) - [Configuration](#configuration) - [Platform Adapters](#platform-adapters) - [LLM Providers](#llm-providers) - [Tools](#tools) - [RAG (Document Search)](#rag-document-search) - [Hub](#hub) - [Memory](#memory) - [Access Control](#access-control) - [Security Notes](#security-notes) - [Contributing](#contributing) - [License](#license) - [Contributors](#contributors) - [Star History](#star-history) --- ## Why Garudust? - **~10 MB binary, < 20 ms cold start** — statically linked, zero runtime dependencies - **Self-improving** — learns your preferences, auto-generates reusable skills, corrects itself without being told twice - **Parallel tool execution** — conflict-aware grouping runs independent tools simultaneously, serializes only when necessary - **Credential rotation** — set `LLM_FALLBACK_API_KEYS` and the agent rotates to the next key on any auth failure, automatically - **Smart context compression** — 3-region strategy preserves the original task and recent turns; only the middle is summarized - **Lifecycle hooks** — `AgentHooks` callbacks for every turn, compression event, delegation, and session end - **agentskills.io compatible** — install community skills from the hub or any GitHub repo with one command - **7 platform adapters** — Telegram, Discord, Slack, Matrix, LINE, WhatsApp, Webhook — all in one process - **24 LLM providers, named profiles** — Anthropic, OpenAI, Gemini, Groq, Mistral, DeepSeek, xAI, Together AI, Fireworks, Cerebras, Perplexity, Cohere, NVIDIA NIM, Alibaba DashScope, ByteDance Doubao, Zhipu AI, Moonshot, Baidu ERNIE, OpenRouter, AWS Bedrock, Ollama, vLLM, ThaiLLM, or any OpenAI-compatible endpoint — configure named `providers:` profiles in `config.yaml` and route per-task - **Provider routing hints** — map hint names to provider/model pairs in config; pass `--hint fast` to route a single task to a cheaper model without changing the default - **Per-tool model config** — override the model (and fallback) used by each hub tool or skill script via `tools.<name>.model` in `config.yaml` - **Secure by design** — Docker sandbox, hardline command blocks, memory-poisoning protection, automatic secret redaction --- ## Supported Platforms Download a pre-built binary from [**GitHub Releases**](https://github.com/garudust-org/garudust-agent/releases/latest) — no Rust required:
anything you tell the agent to stop doing — saved immediately
always-on at `POST /webhook` — no token needed
Description
Run shell commands (Docker sandbox optional — see security note below)
Who
axum HTTP gateway — `/chat`, `/chat/stream`, `/chat/ws`, `/metrics`
Examples
`LINE_CHANNEL_TOKEN`, `LINE_CHANNEL_SECRET`
`vllm` *(add `url:` for custom endpoint)*
Write a file to the filesystem; sensitive credential paths are always blocked
Load a skill's full instructions by name
Telegram, Discord, Slack, Matrix, LINE, WhatsApp, Webhook
`DISCORD_TOKEN`
`openrouter`
Arbitrary HTTP requests with custom headers and body
Full-text search across past conversations (FTS5 trigram)
Tool registry + built-in toolsets (web, files, terminal, browser, …)
deepseek # xai
`GEMINI_API_KEY`
`ZHIPU_API_KEY`
`auto` / `smart` / `deny`
gemini
`openai`
`ARK_API_KEY`
Values
`FIREWORKS_API_KEY`
`mistral`
`QIANFAN_API_KEY`
list of tool names
`deepseek`
`cerebras`
`AWS_ACCESS_KEY_ID` + `AWS_SECRET_ACCESS_KEY`
`NVIDIA_API_KEY`
List all documents indexed in the current session
Generate a shareable invite code (1-use, 24 h expiry by default)
`DASHSCOPE_API_KEY`
Remove one or all documents from the RAG index
vllm
`groq`
`TOGETHER_API_KEY`
list of toolset names
`MOONSHOT_API_KEY`
`XAI_API_KEY`
`perplexity`
Index a document (PDF, TXT, CSV, MD, …) for full-text search
deny # smart = audits risky calls but does NOT block them; # use deny to block all tool use without explicit allow-list terminal_sandbox: none # none
`cohere`
Full-text search across all indexed documents
Show your current platform ID and role
list of tool names
What it involves
filesystem
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.