Minibot — Self-Hosted AI Assistant for Telegram¶
A lightweight Python AI agent with persistent memory, tools, skills, scheduling, MCP, and multi-agent orchestration.
Minibot is an open-source, self-hosted AI assistant for Telegram. It runs on your own infrastructure — Telegram-first, SQLite-first, async-first — so your chats, memory, and scheduled automations stay in your instance, not a third-party SaaS.
pip install minibot
minibot configure # interactive wizard: Telegram token, provider, tools
Quickstart · GitHub · PyPI
What Minibot can do¶
Persistent memory — durable facts and preferences, stored in SQLite.
Agent tools — Python, Bash, HTTP, file storage, grep, and patch-based editing.
Skills and agents — file-based skills plus specialist agents with delegation.
Scheduling — one-shot, interval, and cron prompts for Telegram automations.
Async tasks — offload long-running work to a SQLite- or RabbitMQ-backed worker queue.
MCP integration — connect external Model Context Protocol servers.
RAG — index documents into SQLite (or Qdrant) for retrieval-augmented answers.
Relation graph — optional typed, owner-scoped relationships between entities.
Browser automation — Playwright-driven browsing and screenshots.
Voice — optional speech-to-text for audio messages.
Why self-host¶
Privacy & ownership: transcripts, KV notes, and scheduled prompts live in your instance (SQLite files), not a third-party service.
Cost & provider control: pick where to route LLM calls and manage API usage independently.
Runtime control: deploy behind your firewall, restrict outbound access, and run the daemon as an unprivileged user.
Getting started
Features
Extending MiniBot
For AI agents¶
Machine-readable versions of these docs are published alongside the site:
llms.txt — a concise, curated index for LLM and AI-agent consumers.
llms-full.txt — a single-file dump of the complete documentation.