I've been fascinated by the rise of personal AI agents like OpenClaw and Manus. The idea of an AI that acts on your behalf—searching, scheduling, remembering—is compelling.
But every time I tried one, the setup killed it for me. API keys, local installation, LLM provider config, permissions. By the time it was running, I'd spent an hour and still didn't have recurring tasks working.
So I built Tether AI—a personal AI agent that lives inside Telegram.
How it works
- Sign in with Google at https://www.trytether.ai
- Link your Telegram
- Message @HeyTetherBot like a friend
That's the entire setup. No API keys. No self-hosting. No app to install.
What it can do
- Web search: ask it anything, it searches and summarizes.
- Reminders: "remind me to call mom at 5pm."
- Memory: remembers your preferences across conversations.
- Voice & images: send voice messages, understand images.
- Google Calendar: check and manage your schedule.
- Recurring routines: this is the big one.
Routines: the killer feature
Set up a routine once and it runs forever. Example:
"Every weekday at 7am, send me a briefing with today's calendar, top tech news, and weather in San Francisco."
It runs while you sleep and delivers to your Telegram every morning. No prompting. No opening an app.
This is what separates an agent from a chatbot. If it can't do anything while you're offline, it's just a chatbot.
Transparency
Every action the agent takes is logged in a full activity feed. You can see exactly what tools it used, what it searched, and what it decided. No black box.
Tech stack
- Next.js (App Router)
- Claude (Anthropic) for reasoning
- Telegram Bot API
- Google APIs (Calendar)
- Vercel (hosting + cron jobs)
- PostgreSQL
Try it
- App: https://www.trytether.ai (free to use, no credit card)
- Community: https://t.me/TetherAICommunity
Happy to answer questions about the architecture or product decisions.
Top comments (0)