Control AI coding agents from your phone. CCGram bridges Telegram to tmux — monitor output, respond to prompts, and manage sessions without touching your computer. Supports Claude Code, Codex CLI, Gemini CLI, and plain shell sessions with LLM command generation.
AI coding agents run in your terminal. When you step away — commuting, on the couch, or just away from your desk — the session keeps working, but you lose visibility and control.
CCGram fixes this. The key insight: it operates on tmux, not any agent's SDK. Your agent process stays exactly where it is, in a tmux window on your machine. CCGram reads its output and sends keystrokes to it. This means:
- Desktop to phone, mid-conversation — your agent is working on a refactor? Walk away and keep monitoring from Telegram
- Phone back to desktop, anytime —
tmux attachand you're back in the terminal with full scrollback - Multiple sessions in parallel — Each Telegram topic maps to a separate tmux window, each can run a different agent
Other Telegram bots wrap agent SDKs to create isolated API sessions that can't be resumed in your terminal. CCGram is different — it's a thin control layer over tmux, so the terminal remains the source of truth.
graph LR
subgraph phone["📱 Telegram Group"]
T1["💬 Topic: api"]
T2["💬 Topic: ui"]
T3["💬 Topic: docs"]
T4["💬 Topic: server"]
end
subgraph machine["🖥️ Your Machine — tmux"]
W1["🟠 window @0<br>claude ↻ running"]
W2["🧩 window @1<br>codex ↻ running"]
W3["♊ window @2<br>gemini ↻ running"]
W4["🐚 window @3<br>shell ↻ ready"]
end
T1 -- "text →" --> W1
W1 -. "← responses" .-> T1
T2 -- "text →" --> W2
W2 -. "← responses" .-> T2
T3 -- "text →" --> W3
W3 -. "← responses" .-> T3
T4 -- "text / voice →" --> W4
W4 -. "← output" .-> T4
style phone fill:#e8f4fd,stroke:#0088cc,stroke-width:2px,color:#333
style machine fill:#f0faf0,stroke:#2ea44f,stroke-width:2px,color:#333
style T1 fill:#fff,stroke:#0088cc,stroke-width:1px,color:#333
style T2 fill:#fff,stroke:#0088cc,stroke-width:1px,color:#333
style T3 fill:#fff,stroke:#0088cc,stroke-width:1px,color:#333
style T4 fill:#fff,stroke:#0088cc,stroke-width:1px,color:#333
style W1 fill:#fff,stroke:#2ea44f,stroke-width:1px,color:#333
style W2 fill:#fff,stroke:#2ea44f,stroke-width:1px,color:#333
style W3 fill:#fff,stroke:#2ea44f,stroke-width:1px,color:#333
style W4 fill:#fff,stroke:#2ea44f,stroke-width:1px,color:#333
Each Telegram Forum topic binds to one tmux window running an agent CLI. Messages you type in the topic are sent as keystrokes to the tmux pane; the agent's output is parsed from session transcripts and delivered back as Telegram messages.
Session control
- Send messages directly to your agent topic
/commandsshows commands supported by that topic's provider (Claude/Codex/Gemini/Shell)- Forwarded slash commands report provider mismatch errors (for example Claude-only
/costin Codex) - Command menu auto-switches per user/chat to the active topic provider after interaction
- Interactive prompts (AskUserQuestion, ExitPlanMode, Permission) rendered as inline keyboards
- Codex edit approvals are reformatted for Telegram readability (compact summary + short preview, with approval choices preserved)
- Codex
/statusreplies include a bot-side transcript snapshot (session + token/rate-limit stats) when Codex does not emit a normal transcript message - Multi-pane support — auto-detects blocked panes, surfaces prompts,
/panescommand for overview - Terminal screenshots — capture the current pane (or any specific pane) as a PNG image
- Voice message transcription via Whisper API (OpenAI, Groq) with confirm/discard keyboard
- Sessions dashboard (
/sessions) — overview of all sessions with status and kill buttons - Remote Control detection — 📡 topic badge when RC is active, one-tap activation from status keyboard
- Action toolbar (
/toolbar) — persistent inline buttons for RC, Screenshot, Esc, Notify, Ctrl-C
Real-time monitoring
- Assistant responses, thinking content, tool use/result pairs, and command output
- Live status line showing what the agent is currently doing
- Entity-based formatting with automatic plain text fallback
Session management
- Directory browser for creating new sessions from Telegram
- Auto-sync: create a tmux window manually and the bot auto-creates a matching topic
- Fresh/Continue/Resume recovery when a session dies
- Message history with paginated browsing (
/history) - Persistent state — bindings and read offsets survive restarts
Multi-provider support
- Claude Code (default), OpenAI Codex CLI, Google Gemini CLI, and Shell
- Per-topic provider selection — different topics can use different agents simultaneously
- Auto-detects provider from externally created tmux windows (process name, with ps-based TTY detection fallback for JS-runtime-wrapped CLIs like bun/node)
- Provider-aware recovery (Continue/Resume buttons adapt to each provider's capabilities)
- Emdash integration — auto-discovers emdash tmux sessions; bind Telegram topics to emdash-managed agents with zero configuration
Shell provider
- Chat-first: type natural language → LLM generates a shell command → approve with one tap → output streams back
- Raw mode: prefix with
!to bypass the LLM and send commands directly - Voice-to-command: voice messages transcribed via Whisper, then routed through the LLM for command generation
- Dangerous command detection with extra confirmation step
- BYOK LLM — supports OpenAI, Anthropic, xAI, DeepSeek, Groq, Ollama (zero new dependencies)
- See docs/providers.md for LLM configuration
Inter-agent messaging (swarm)
- Agents discover each other, exchange messages, broadcast notifications, and spawn new agents
- File-based mailbox (
~/.ccgram/mailbox/) — no database, no daemon - Broker delivers pending messages to idle agent windows automatically
- Telegram visibility — silent notifications in sender/recipient topics
- Spawn approval — agents request new instances, user approves via Telegram
- See docs/guides.md for setup and usage
Extensibility
- Global Telegram menu includes bot commands + default provider commands (with
↗prefix); provider-scoped menus auto-refresh per chat/user/topic context with Telegram-scope fallbacks - Tmux session auto-detection — when running inside tmux, auto-discovers the session and picks up existing agent windows; duplicate instance prevention
- Multi-instance support — run separate bots per Telegram group on the same machine
- Configurable via environment variables
- Python 3.14+
- tmux — installed and in PATH
- At least one agent CLI —
claude(default),codex, orgeminiinstalled and authenticated (or use the plainshellprovider with no extra install required)
# Recommended
uv tool install ccgram
# Alternatives
pipx install ccgram # pipx
brew install alexei-led/tap/ccgram # Homebrew (macOS)- Create a Telegram bot via @BotFather
- Configure bot settings in BotFather:
- Allow Groups: Enabled (Bot Settings > Groups & Channels > Allow Groups? > Turn on)
- Group Privacy: Disabled (Bot Settings > Groups & Channels > Group Privacy > Turn off) — Required to see all messages in topics
- Topics: Enabled (Bot Settings > Groups & Channels > Edit Topics > Enable)
- Add the bot to a Telegram group with Topics enabled.
- Promote the bot to Administrator and ensure it has these permissions:
- Create Topics (required for the bot to automatically sync and manage session topics)
- Pin Messages (required for the General-topic usage hint)
- Note: the bot also uses message reactions in the General topic — this requires no extra permission.
- Create
~/.ccgram/.env:
TELEGRAM_BOT_TOKEN=your_bot_token_here
ALLOWED_USERS=your_telegram_user_id
CCGRAM_GROUP_ID=your_telegram_group_idGet your user ID from @userinfobot on Telegram. Get the group ID by adding -100 in front of the Peer ID found in the Group Info (or use @RawDataBot).
ccgram hook --installThis registers Claude Code hooks (SessionStart, Notification, Stop, StopFailure, SessionEnd, SubagentStart, SubagentStop, TeammateIdle, TaskCompleted) for automatic session tracking, instant interactive UI detection, API error alerting, session lifecycle cleanup, and agent team notifications. Not needed for Codex or Gemini — those providers are discovered from hookless transcripts and tmux window/provider detection.
If hooks are missing, ccgram warns at startup with the fix command. Hooks are optional — terminal scraping works as fallback.
To enable natural language → shell command generation in shell topics, add an LLM provider:
CCGRAM_LLM_PROVIDER=openai # or: anthropic, xai, deepseek, groq, ollamaSee docs/providers.md for all options. Without an LLM, shell topics forward text as raw commands.
ccgramOpen your Telegram group, create a new topic, send a message — a directory browser appears. Pick a project directory, choose your agent (Claude, Codex, Gemini, or Shell), then choose session mode (✅ Standard or 🚀 YOLO), and you're connected.
CCGram was previously named ccbot. If upgrading from v1.x:
# Install new package
pip install ccgram # or: brew install alexei-led/tap/ccgram
# Migrate config directory
mv ~/.ccbot ~/.ccgram
# Update environment variables: CCBOT_* → CCGRAM_*
# Old CCBOT_* vars still work as fallback with deprecation warnings
# Re-install hooks (replaces legacy "ccbot hook" entries)
ccgram hook --install- docs/guides.md — CLI reference, configuration, upgrading, multi-instance setup, session recovery, testing
- docs/providers.md — Provider details (Claude, Codex, Gemini, Shell), session modes, LLM configuration, custom launch commands
git clone https://github.com/alexei-led/ccgram.git
cd ccgram
uv sync --extra dev
make check # fmt + lint + typecheck + unit + integration tests
make test-e2e # E2E tests (requires agent CLIs, see docs/guides.md)CCGram started as a fork of ccbot by six-ddc, who created the original Telegram-to-Claude-Code bridge. This project has since been rewritten and developed independently with multi-provider support, topic-based architecture, interactive UI, and a comprehensive test suite. Thanks to six-ddc for the initial idea and implementation.