Skillget

tuskbot

mcp
SUMMARY

Tusk — Personal AI Assistant

README.md

TuskBot 🦣

Build
Go Report
Release
Go Version
License

Autonomous AI Agent

TuskBot is a Go-based autonomous agent designed to handle tasks directly in your Telegram. It evolves the ideas of OpenClaw into a more scalable, secure, tool-oriented assistant.

🎯 Motivation

TuskBot originated as an evolution of the OpenClaw concept, addressing its core architectural limitations:

  • High-Performance Engine: Built with Go for native concurrency and high-speed execution, no interpreted overheads.
  • Modular Stability: Powered by Model Context Protocol (MCP). Tool isolation prevents cascading failures and ensures system resilience.
  • Persistent Context: Full Local RAG pipeline (SQLite-vec + llama.cpp). No need to send all your chats to OpenAI for embedding.
  • Privacy-First Design: Native support for Ollama and local embedding models.

🚀 Capabilities

🔌 Extensible via MCP

TuskBot uses a Model Context Protocol (MCP)-first approach. This allows you to plug in any MCP-compliant server (databases, APIs, or local tools) without modifying the core logic. If a tool exists as an MCP server, TuskBot can use it.

🧠 Private RAG & Persistent Memory

The bot maintains a long-term memory of your interactions using a local Retrieval-Augmented Generation (RAG) pipeline:

  • Zero-API Embeddings: Uses embedded llama.cpp (via GGUF models) to process text locally. Your data for semantic search never leaves your hardware.
  • Vector Storage: Powered by SQLite-vec for fast, local retrieval of conversation history and technical context.

🛠️ System Access

TuskBot comes with a set of pre-configured tools for immediate use:

  • Filesystem: Manage, read, and write files in the bot's workspace.
  • Shell Execution: Run system commands and scripts directly through the chat.
  • MCP Manager: Allows agent to connect and restart MCP servers.

💾 Installation

Download the pre-compiled binary for your platform from the Releases page.

Quick Install (Linux/macOS):

curl -fsSL https://raw.githubusercontent.com/sandevgo/tuskbot/main/scripts/install.sh | sh

Supported release artifacts:

  • Linux amd64
  • Linux arm64
  • macOS arm64

The installer downloads the latest stable release binary, installs it into a user-local bin directory, then runs:

  • tusk install (interactive setup)
  • tusk service install
  • On Linux user-mode installs (TUSK_SERVICE_USER_MODE=true), attempts loginctl enable-linger <user> (may require sudo)
  • tusk service start

Manual install fallback:

# Example for Linux amd64 (replace filename for your platform)
tar -xzvf tusk-linux-amd64.tar.gz
chmod +x bin/tusk-linux-amd64
mkdir -p ~/.local/bin
mv bin/tusk-linux-amd64 ~/.local/bin/tusk

Running TuskBot

tusk run

Using Docker

Docker compose example:

services:
  tuskbot:
    image: ghcr.io/sandevgo/tuskbot:latest
    volumes:
      - tuskbot-data:/root/.tuskbot
    command: run

volumes:
  tuskbot-data:

Running installation with Docker Compose:

# Configure
docker compose run tuskbot install

# Run
docker compose up -d

⌨️ Slash Commands

TuskBot supports the following slash commands for direct interaction:

  • /model Display/Switch the currently active LLM provider and model.
  • /mcp List all currently connected MCP servers and their available tools.

🔧 Configuration

TuskBot uses environment variables for configuration.

Core Settings

  • TUSK_TELEGRAM_TOKEN: Your Telegram Bot Token.
  • TUSK_TELEGRAM_OWNER_ID: Your Telegram User ID (for security).
  • TUSK_CHAT_CHANNEL: Primary chat interface (e.g., telegram).
  • TUSK_RUNTIME_PATH: Path for logs, database, and workspace (default: ~/.tuskbot).
  • TUSK_DEBUG: Enable debug logging (set to 1).

AI & Memory

  • TUSK_MAIN_MODEL: Main LLM model (format: provider/model).
  • TUSK_EMBEDDING_MODEL: Embedding model file name (gguf).
  • TUSK_CONTEXT_WINDOW_SIZE: Number of messages in active context (default: 30).

Providers

  • TUSK_OPENROUTER_API_KEY: API Key for OpenRouter.
  • TUSK_OPENAI_API_KEY: API Key for OpenAI.
  • TUSK_ANTHROPIC_API_KEY: API Key for Anthropic.
  • TUSK_OLLAMA_BASE_URL: Base URL for Ollama (default: http://127.0.0.1:11434).
  • TUSK_OLLAMA_API_KEY: API Key for Ollama (optional).
  • TUSK_CUSTOM_OPENAI_BASE_URL: Base URL for Custom OpenAI provider.
  • TUSK_CUSTOM_OPENAI_API_KEY: API Key for Custom OpenAI provider.

System Service

  • TUSK_SERVICE_USER_MODE: Install service in user mode by default (default: true).
  • TUSK_SERVICE_LOG_DIRECTORY: Directory for generated service stdout/stderr log files (default: TUSK_RUNTIME_PATH).

🗺 Roadmap

  • [X] Unified Command Interface: Support of slash-commands (/).
  • [X] Cron/heartbeat: Scheduled tasks and periodic checks.
  • [X] Multi-Agent Orchestration: Sub-agents to delegate specialized tasks
  • [ ] Skills: Skills for agents to perform specific actions.

Reviews (0)

No results found