firetg

Telegram MTProto CLI for scripts and agents, powered by mtcute.

Install

Run directly:

bunx firetg --help

Or install globally:

bun install -g firetg
firetg --help

Agent skill

Install the firetg skill through skills.sh:

bunx skills add Mergemat/firetg --skill firetg

The skill teaches agents to bound reads, keep authentication secrets in the terminal, and confirm recipients and content before sending messages.

Login

Create a Telegram app at https://my.telegram.org/apps, then run:

firetg auth login

The CLI asks for your API ID and API hash, then shows a QR code to scan in Telegram.

Phone login is also supported:

firetg auth login --phone

Credentials are stored in ~/.config/firetg/config.json. Telegram auth state and the peer cache are stored in ~/.config/firetg/telegram.sqlite.

Existing Teleproto/GramJS string sessions at ~/.config/firetg/session are converted to mtcute storage on the first authenticated command. The legacy session and peers.json files are removed after a successful conversion.

Set XDG_CONFIG_HOME to place the firetg directory elsewhere. If it is not set, firetg uses the current user's standard ~/.config directory.

Commands

firetg status --json --pretty
firetg doctor --json --no-input --timeout 15

firetg profiles me
firetg profiles get telegram
firetg profiles get 123456789

firetg channels view --username telegram
firetg channels view --id 100
firetg channels messages --username example_channel --limit 50
firetg channels pinned --username example_channel --limit 20

firetg messages send --username telegram --text "hello"
firetg messages send --id 123456789 --text "hello"
firetg messages send --username telegram --file ./photo.jpg --text "caption"
firetg messages send --username telegram --file ./report.pdf --document
firetg messages send --username telegram --text "hello later" --schedule-at 2030-07-05T15:00
firetg messages list --chat me --limit 20
firetg messages list --chats alice,bob,me --limit 20
firetg messages list --chat me --search deploy --limit 10
firetg messages search --chat launch-team --hashtag "#deploy" --limit 100
firetg messages search --chat launch-team --reply-to 101 --from 42,alice --limit 50

firetg folders list
firetg dialogs list --folder 1 --limit 20

firetg auth logout

For batch history, --chats alice,bob,me --limit 20 reads 20 messages per dialog on one connection. Results are grouped by chat with individual errors; successful reads remain available if another chat fails.

All commands accept global agent controls:

firetg messages list --chat me --limit 50 --no-input --timeout 30
firetg messages list --chat me --limit 50 --output /tmp/messages.json

--pretty formats JSON for people. --output <path> keeps stdout empty and writes the result with file mode 0600; its confirmation goes to stderr. --no-input fails rather than prompting, and --timeout <seconds> returns a structured TIMEOUT error with exit code 2.

Use --help for more detail:

firetg --help
firetg messages --help
firetg channels view --help
firetg channels messages --help
firetg channels pinned --help
firetg messages list --help
firetg messages search --help

Output

Successful commands print JSON to stdout. Telegram, configuration, and rate-limit failures retain structured JSON when agents need to branch or schedule a retry. Command/argument mistakes print concise text plus relevant usage so an agent does not need a second --help call. Prompts and output-file confirmations print to stderr.

Success output is the command result itself:

{}

Error:

{
  "ok": false,
  "error": {
    "code": "CONFIG_ERROR",
    "message": "Missing config file at /path/to/config.json"
  }
}

Usage error:

Unknown command: dialogs listdd.

Available dialogs commands:
  firetg dialogs list [--folder <id>] [--limit <n>]

Message-reading commands return at most 100 items. Text is limited to a 1,000-character preview by default and includes "textTruncated":true when shortened. Pass --full-text only when complete bodies are needed.

Contributing

See CONTRIBUTING.md for development and pull request guidelines. Report security vulnerabilities according to SECURITY.md.

License

Released under the GPL-3.0-or-later license.