aipager 0.2.0

Telegram remote-control daemon for Claude Code CLI sessions running in dtach

aipager

Telegram remote-control for Claude Code CLI sessions. Run Claude inside a detached terminal (dtach), drive it from your phone — read responses, send prompts, approve permission requests, switch sessions — without an SSH session staying open.

Install

Requires Python 3.10+. dtach is installed automatically via the dtach-bin dependency — no separate system package needed.

pipx install aipager

(or uv tool install aipager, or pip install aipager into a venv — all work the same.)

Linux ARM and macOS users on Apple Silicon get the same one-command install; the dtach binary ships as pre-built wheels for each platform.

Homebrew support is coming in v0.3 as brew install <user>/tap/aipager, which uses the system dtach instead of the bundled one.

Configure

aipager config

Interactive wizard — asks for your Telegram bot token (from @BotFather) and chat ID, validates them, then patches ~/.claude/settings.json to wire the necessary hooks automatically. You never edit any file by hand.

Run

aipager start

The daemon stays in the foreground. Launch a Claude session in another terminal:

claude-dtach dev

The daemon discovers the session within seconds and Telegram starts mirroring it. To survive logout, use screen, tmux, or a systemd-user unit (template at scripts/aipager.service.example — full aipager service install automation lands in v0.4).

What it does

• Mirrors Claude Code session state to Telegram: busy/idle, tool calls, context %, cost, line counts
• Lets you reply to messages to inject prompts back into the session
• Surfaces permission prompts and AskUserQuestion dialogs as Telegram inline keyboards
• Notifies on context warnings, compaction, session end, and stalls
• Supports multiple concurrent sessions with one bot
• Optional read-only observer bots

Developing locally

git clone <repo-url> aipager && cd aipager
python3 -m venv .venv && source .venv/bin/activate
pip install -e '.[dev]'
pytest -q

When iterating on code changes you'll generally want to also install dtach-bin from a local checkout — or pip install dtach-bin — so the runtime can find dtach on PATH.

Release process is in CONTRIBUTING.md.

License

MIT — see LICENSE.