⌘ Free · Runs entirely on your Mac

Your own AI,
living in your chat

Telebot AI bridges your Telegram bots to any LLM — local models running on your Mac, or cloud AI from OpenAI, Anthropic, Groq and more. One bridge, every chat, entirely yours.

Telebot AI app icon
Download for macOS
See what it does

macOS 13+ · Apple Silicon · requirements

Telebot AI — Telegram ⋯
remind me to water the plants at 6pm 🌱
Telebot AI · just now Done — watering plants 🌱 at 18:00. I'll ping you when it's time.
🔔
Water the plants today · 18:00
✓
/dice
Telebot AI · just now
You rolled a 3 🎲
Would an M5 Mac beat an RTX 4090 for local AI?
Telebot AI · just now Short version: depends on the job —
M5 + MLX 128GB unified memory — fits models a 4090 can't even load
RTX 4090 faster tokens, but capped at 24GB VRAM
Call M5 wins for big models & memory — 4090 for pure speed

Bring your own brain — local models or any cloud AI

mlx-serve Ollama LM Studio vLLM OpenAI Anthropic Groq OpenRouter

Features

Everything your assistant
should just do.

🧰

30+ tools, model-driven

Reminders, weather, stocks, crypto, polls, dice, expenses, habits, bookmarks, RSS feeds, translation, currency, PDFs & receipt scanning, memory, digests and briefings — the model picks the tool, no commands to memorize. Extend it with any MCP server.

/remind /weather /stocks /poll /roll /translate /briefing
🌐

Local or cloud

Run fully offline with mlx-serve or Ollama on your own Mac — or plug in any cloud AI: OpenAI, Anthropic, Groq, OpenRouter. Stack several as automatic fallbacks so the bot keeps talking even if one goes down.

💬

Telegram-native

Streaming replies, typing dots, animated dice, native polls, voice notes in and out. Feels like chatting with a friend — not a dashboard.

🔒

Local-first privacy

Conversation history, memory and skills live in ~/.mlx-serve on your Mac. Nothing leaves your machine unless you point it at a remote model.

🗝️

Keyless data sources

Weather & air quality, currency rates (ECB), Wikipedia, SEC filings, RSS — all fetched with zero API keys. No sign-ups, no rate-limit dashboards.

👥

Multi-bot mode

One token per bot. Run several assistants with different models, personalities and skills — side by side, from a single Mac.

⚡

Agentic extras

Price alerts, nightly auto-memory with pruning, weekly reviews and a morning briefing at 08:00. Your assistant keeps working while you sleep.

🍱

Menu bar control

Start, stop, restart, open settings or tail the logs — all from the menu bar. The bot runs as a LaunchAgent, always on, out of your way.

🗨️

DMs and groups

One bot for private chats and groups. In a group it only answers when mentioned or replied to — and uses the surrounding conversation as context, so it follows the topic without spamming the chat.

📚

Skill playbooks

Skills are playbooks the bot follows — briefings, history search and more. Toggle them per bot from the skill registry, and each one speaks your language. No code, no plugins to install.

🧠

Searchable memory

It remembers your chats — and you can search them. Ask ‘what did we say about that?’ and get dated snippets from months ago, per chat, in private chats and groups.

🌍

19 languages

The app, its settings, and this website. Switch from the menu bar, the settings window, or the language menu on any page.

English 简体中文 Español 日本語 Deutsch Français Português 한국어 Italiano Русский العربية हिन्दी Bahasa Indonesia Türkçe Nederlands Polski Tiếng Việt ไทย Українська

Getting started

Up and running in two minutes.

01

Download the .dmg

Open it and drag Telebot AI into Applications. The build isn't notarized yet, so if macOS shows a warning, right-click (Control-click) the app and choose Open — once, and it launches normally from then on.

02

Create your bot

Message @BotFather on Telegram, get a token, and paste it into the settings window. One token per bot — run as many as you like.

03

Connect a brain

Point it at mlx-serve, Ollama, LM Studio — or any API provider. Add fallbacks and the bot keeps talking even if one model goes down. New to local models? Read the guide.

Privacy

Your conversations never
leave your Mac.

No cloud account. No telemetry. No training on your chats. Every message, memory and skill lives in one folder you own.

everything lives in ~/.mlx-serve/

Requirements

Built for the Mac.

Hardware & OS

  • macOS 13 (Ventura) or later
  • Apple Silicon — M1 or newer
  • Local models like RAM: the bigger the better
  • No GPU needed — MLX uses the Neural Engine

Software

  • A Telegram account & bot token from @BotFather
  • Any OpenAI-compatible backend — local or cloud
  • Optional: mlx-serve, Ollama or LM Studio for fully offline AI

Note: current builds are unsigned — on first launch use right-click → Open, or remove the quarantine flag with xattr -dr com.apple.quarantine /Applications/Telebot\ AI.app. Signed & notarized builds are on the way.

FAQ

Answers before you ask.

Do I need an API key or subscription?

No. Run a local model (mlx-serve, Ollama, LM Studio) and everything is free. Data sources — weather, currency, Wikipedia, SEC filings — need no keys. Cloud APIs (OpenAI, Anthropic, Groq, OpenRouter…) are optional and bring-your-own.

Which models can I use?

Anything that speaks OpenAI-compatible HTTP — from tiny local models on your Mac to frontier cloud models like OpenAI and Anthropic Claude. The settings app lets you add fallback providers, so the bot keeps answering even when one is down.

Does my data leave my Mac?

Your chat history, memories and skills stay in ~/.mlx-serve. The only outbound traffic is the Telegram API itself, whatever model server you choose to point the bot at, and anonymous usage stats.

Do I need a powerful Mac?

The menu bar app itself is tiny. Local models scale with your RAM — a 16 GB M-series Mac runs small models comfortably. No Mac? Point the bot at any cloud API instead.

Does it work on Intel Macs?

Not yet — the current build is Apple Silicon only (M1 or newer). The unified-memory architecture is what lets it run big local models; an Intel build is on the roadmap.

Can I run multiple bots?

Yes. Create a bot per token with @BotFather, add them all in settings, and give each one its own model, personality file and skills.

How do I update the app?

The app checks for new versions automatically — the menu bar item and the Updates page in settings will tell you when a new build is out and link straight to the download. Or grab the latest .dmg from this page and drag it over the old app in Applications. Your settings, bots and memories live outside the app, so they carry over untouched.

Why does macOS warn about the app?

Because the build isn't notarized by Apple yet — that costs money and requires a paid developer account. The warning is expected: right-click (Control-click) the app in Applications, choose Open, then Open again. One time, and it runs normally from then on.

Is it really free?

Yes — free to download and use. No subscriptions, no accounts. You only pay for the model provider you choose, if any.

Give your Mac a brain.
Put it in your pocket.

Download the .dmg, drag it to Applications, and say hi to your new assistant.

Telebot AI app icon
Download Telebot AI
40.0 MB · Apple Silicon · SHA-256

macOS 13+ · requirements · FAQ

☕ Buy me a coffee

Feedback

Tell us what you think.