takopi

🐙 he just wants to help-pi

telegram bot for codex. with streaming progress and resumeable sessions.

features

stateless resume via codex resume <token> lines in chat.

edits progress message while codex runs (commands, tools, reasoning, edits, elapsed time).

renders markdown to telegram entities.

runs in parallel across threads and queues per thread to keep codex history sane.

requirements

• uv for installation (curl -LsSf https://astral.sh/uv/install.sh | sh)
• codex on PATH (npm install -g @openai/codex or brew install codex)

install

• uv tool install takopi to install as takopi
• or try it out with uvx takopi

setup

1. get bot_token from @BotFather
2. get chat_id from @myidbot
3. send /start to the bot (telegram won't let it message you first)
4. run codex once interactively in the repo to trust the directory

config

takopi config lives in ~/.takopi/takopi.toml and can be overriden by .takopi/takopi.toml in the current repo.

bot_token = "123456789:ABCdefGHIjklMNOpqrsTUVwxyz"
chat_id = 123456789

[codex]
# optional: profile from ~/.codex/config.toml
profile = "takopi"

usage

start takopi in the repo you want to work on:

cd ~/dev/your-repo
takopi

send a message to the bot.

to continue a thread, reply to a bot message.

to stop a run, reply to the progress message with /cancel.

cli

by default takopi sends the progress message silently, and replaces it with the final answer to notify you. this can be turned off with --no-final-notify

development

see docs/specification.md and docs/developing.md.