Token Beggar 0.3.0

Token Beggar 0.3.0 compares two setups: Without Token Beggar (the chat model alone) and With Token Beggar (this release).

What changed

• The card now sizes the task first. Small tasks stay in the chat. A large task, such as fixing a bug in code the model has not read, goes to one cheaper worker, and the chat reads the whole diff and runs the tests before it answers.
• Mode is chosen from the chat model and effort (full, sized, off). token-beggar doctor lists what can quietly stop routing.
• The footer shows a quota multiple and estimated dollars saved, never token counts, and no multiple under 1.15x.

Results (Opus 5.5 at high effort, list price, 18 real bug fixes from 12 Python projects, scored by the maintainers' hidden tests, 3 runs per task, 54 per setup)

• Without Token Beggar: $0.69 per task, 70% of issues resolved.
• With Token Beggar: $0.49 per task, 1.42x less (interval 1.33 to 1.51), 69% resolved, a tie within the spread.

Limits

• A blind review by a small model rated the 0.3.0 diffs lower than working alone (0.94 against 1.33 on a 0 to 2 scale), though it agrees only weakly with the tests. It is not shown that the code is as good.
• Measured for one chat model and effort and for Python projects only.
• Small tasks were measured only with an earlier card (1.11x, under the 1.15x line), so expect close to no change on small one-file edits.

Install: claude plugin marketplace add Arthur031221/Token-Beggar then claude plugin install token-beggar@token-beggar. Details in docs/benchmark.md.