Skip to main content
Easiest channel to get started. Telegram is the fastest way to give Fliiq its own communication channel — just create a bot via BotFather, no app passwords, no paid services.
Fliiq gets its own Telegram bot. This isn’t your personal Telegram — it’s Fliiq’s. You message Fliiq’s bot from your Telegram app the same way you’d message any person or bot. Your Telegram talks to Fliiq’s Telegram. Fliiq integrates with Telegram in two ways: outbound messaging via send_telegram skill, and a real-time listener in the daemon that responds to incoming messages.

​
Setup

1

Create Fliiq's bot

Message @BotFather on Telegram and use /newbot. BotFather gives you a token like 123456:ABC-DEF....
2

Add the bot token to .env

.env
TELEGRAM_BOT_TOKEN=123456:ABC-DEF...
3

Run the interactive setup

fliiq telegram setup
This will prompt you to send a message to your bot, then auto-detect and save your chat ID to TELEGRAM_ALLOWED_CHAT_IDS in your .env.Manual alternative: Add TELEGRAM_ALLOWED_CHAT_IDS=<chat_id> to ~/.fliiq/.env directly. To find your chat ID, message the bot and check https://api.telegram.org/bot<TOKEN>/getUpdates. Comma-separate multiple IDs.
TELEGRAM_ALLOWED_CHAT_IDS is required. When TELEGRAM_BOT_TOKEN is set, TELEGRAM_ALLOWED_CHAT_IDS must also be set. The daemon will refuse to start without it — this prevents unauthorized users from accessing your agent.

​
Outbound: send_telegram

Send a message from any Fliiq session: 
"Send a Telegram message to chat 12345678 saying the build passed"
Works in interactive sessions, single-shot commands, and scheduled jobs.

​
Real-Time Listener

When the daemon starts and TELEGRAM_BOT_TOKEN is set, Fliiq automatically runs a Telegram listener that:
  1. Long-polls Telegram’s getUpdates API for new messages
  2. Maintains a persistent session store — each chat gets its own conversation history that survives job executions and daemon restarts. Context is never lost between interactions.
  3. Runs a headless agent_loop() for each incoming message
  4. Shows a persistent typing indicator throughout processing — stays active for long-running tasks so you always know the agent is working
  5. Splits long responses to respect Telegram’s message length limits
  6. Filters by allowed chat IDs for security
fliiq daemon start       # Starts daemon + Telegram listener
fliiq daemon status      # Shows "Telegram: active" if connected

​
Using It

Once the daemon is running, message Fliiq’s bot from your phone or Telegram app. The bot responds using the full Fliiq agent loop — same skills, same memory, same capabilities as the CLI. It’s like texting a colleague who happens to have access to your codebase, email, and scheduled jobs.

​
Forum Topics

Telegram groups can enable Forum Topics — a threaded mode where messages go to named topics (like Slack channels). Fliiq supports this via the message_thread_id parameter on send_telegram and send_telegram_audio. To find a topic’s thread ID, send a message to the topic and check https://api.telegram.org/bot<TOKEN>/getUpdates — look for message_thread_id in the update. 
"Send a Telegram message to chat 12345678, thread 42, saying the release passed"
In a job, pass message_thread_id in the prompt or use send_telegram directly: 
name: release-notifications
trigger:
  type: webhook
prompt: "Send a Telegram message to chat 12345678, thread 42, with the release summary"
skills:
  - send_telegram
delivery:
  type: telegram
  chat_id: "12345678"

​
Per-Request Model Override

Switch models mid-conversation without restarting the daemon. Useful when you want Opus for a complex task and Sonnet for quick lookups — all in the same bot session. 
/model opus-4.6                         # Switch to Opus for this and future messages
/model sonnet-4.5 summarize my inbox    # One-shot: use Sonnet for this message only
use the model gpt-4.1 to write a plan   # Natural language syntax also works
/model reset                            # Revert to the default model
Model aliases from ~/.fliiq/models.yaml are resolved automatically. To confirm the active model, just ask: “what model are you using?”

​
Usage in Jobs

name: daily-standup
trigger:
  type: cron
  schedule: "0 9 * * 1-5"
prompt: "Summarize my git activity from yesterday"
delivery:
  type: telegram
  chat_id: "12345678"