Token Beggar is a plugin for a terminal coding agent (the agent with a plugin marketplace and Agent tool). It needs that agent and Python 3.10 or newer. It uses the standard library only.
/plugin marketplace add Arthur031221/Token-Beggar
/plugin install token-beggar@token-beggar
Restart the agent or run /reload-plugins. The next non-trivial prompt of a session carries the menu.
git clone https://github.com/Arthur031221/Token-Beggar
<your agent command> --plugin-dir ./Token-Beggar
Run:
token-beggar doctor
It prints one line per check and exits 1 if something is broken. A warning is not a failure, but each one names something that would change routing: a forced worker model, a hook error in the log, or a catalog older than 30 days. After a few sessions, token-beggar stats shows the workers that ran. To see the menu the model gets, run token-beggar menu.
The plugin puts token-beggar on the PATH of the agent's shell. Outside the agent, run python3 -m token_beggar menu from the repository root.
- For one shell: start the agent with
TOKEN_BEGGAR=offin its environment. - For good: create a file named
offin the state directory (token-beggar envprints its path). Delete the file to turn it back on. - Remove it:
/plugin uninstall token-beggar
Everything lives in one state directory, printed by token-beggar env: the model inventory, quota marks, caps, notes, the usage ledger and an errors.log. Nothing is sent anywhere. The only network call is the weekly download of the model vendor's public models page.