| name | tollbooth |
| description | Estimate the token count and dollar cost of a prompt or context before sending it. Use to size a request, compare model costs side by side, or sanity-check a budget. Counts tokens with tiktoken when available, else a deterministic heuristic, and prices the request across a small editable table of current-ish models. |
tollbooth
know the bill before the bill.
tollbooth tells you what a prompt will cost before you spend it. Feed it text,
it counts the input tokens and prints a per-model table of estimated dollars —
input-only by default, or include an assumed completion length to see the full
round trip.
What it does
- Counts input tokens (tiktoken if installed, otherwise a deterministic, monotonic heuristic).
- Prices the request across a handful of models — Claude (Opus / Sonnet / Haiku) and OpenAI GPT (4o / 4o-mini).
- Lets you restrict to one model, assume N output tokens, dump JSON, or print just the token count.
Prices live in an editable PRICES table in tollbooth.py and are
approximate — adjust them when providers move the goalposts.
Example
$ echo "Summarize this 12-page contract and flag risky clauses." | tollbooth.py --out 600
13 input tokens, 600 output
model input output total
claude-opus-4-8 $0.0002 $0.0450 $0.0452
claude-sonnet-4-6 $0.0000 $0.0090 $0.0090
claude-haiku-4-5 $0.0000 $0.0024 $0.0024
gpt-4o $0.0000 $0.0060 $0.0060
gpt-4o-mini $0.0000 $0.0004 $0.0004
Output
Deliver the token count and the per-model cost table, not a narration of running
tollbooth.py. Don't explain the tool; show the numbers and stop. No preamble,
no closing take on which model to pick unless asked.
Companion tool
tollbooth.py in this folder does the counting and the arithmetic. Pipe a
prompt or context file through it whenever you want the bill before the bill.
