Tether Tether

Your AI agents, in your pocket.

Claude Code and Codex get stuck without input. Unstick them from Telegram - no SSH, no port forwarding, no exposed dashboards.

๐Ÿ”’ Local-first ๐Ÿ’ฌ Telegram-first ๐Ÿค– Multi-agent Apache 2.0
# How it works
1. Start Tether on your machine
2. Attach your Claude Code / Codex sessions
3. Each session = a topic in your Telegram group
4. Agent gets stuck โ†’ you get a notification
5. Reply in Telegram โ†’ agent continues

Why Tether?

Your AI agents run on your machine. Tether puts them in your Telegram - one group, one topic per agent, notifications when they need you.

Feature SSH Web UI Tether
Works from phone โœ— ~ โœ“
Native notifications โœ— โœ— โœ“
No network config โœ— โœ— โœ“
Multiple agents organized โœ— โœ“ โœ“
Code stays local โœ“ โœ“ โœ“

Telegram-first

Every agent session becomes a topic in a Telegram group. Get notified when an agent needs input. Reply from anywhere.

Local-first

Your code never leaves your machine. No cloud processing, no data collection, no vendor lock-in.

Multi-agent

Claude Code, Codex, and more. All your agents in one Telegram group, each in its own topic.

Web UI included too

Tether also ships a web dashboard on localhost. Useful when you're at your desk, but Telegram is the primary interface - no ports to expose, no VPN to configure.

Git diff viewer
Session persistence
Multi-agent overview

Get started in 2 minutes

Self-hosted, open source, Apache 2.0 licensed. Your agents, your machine, your Telegram.

Quick start

# Clone and run
git clone https://github.com/xithing/tether
cd tether
make install && make start
# Set your Telegram bot token in .env
# Add the bot to a group with topics enabled
  • Works with Claude Code, Codex, and other terminal agents
  • Each agent session = a Telegram topic
  • Native notifications when agents need input
  • Web UI included for desktop use
  • Session persistence across restarts
  • Zero data collection - your code stays local

Want updates on new features?