Your agent sends the same
tool list every call.
We collapse it to 3 tokens.

From call 2 onward, Promptolian replaces your entire tool schema block with TOOLS:[name1,name2] — 97% smaller. No other tool does this. Works on any LLM API.

See the tool schema demo → Calculate your savings API docs

Everyday user

No setup. Free forever.

Install the browser extension and click Compress before sending. Works on ChatGPT, Claude, Gemini and more. Your prompts never leave your device.

Install extension →

Agent developer

97% tool schema · 52.9% context.

One API call per session to cache your schemas. Turn 2 onward costs 3 tokens instead of hundreds. Stacks on top of provider-level prompt caching.

Read the API docs →

Live demo

See Promptolian in action

Type any prompt below and see how much shorter it gets — instantly, in your browser. Try the different compression levels to see how they compare.

Your prompt

Promptolian output

—

Original tokens

—

Compressed tokens

—

Saved

≥98%

Fact preservation

Tier:

Example:

Standard Runs entirely in your browser — instant, private, no data sent anywhere. ~15–20% shorter.

52.9%

History compression

KV-cache sandwich layout across a 20-turn session. The gap widens every turn — by turn 20 you pay for 663 tokens instead of 1,850.

97%

Tool schema — turn 2+

The full JSON block collapses to TOOLS:[name1,name2] on every subsequent call. No other tool does this.

100%

Fact preservation

Every number, entity, file path and named term survives unchanged — verified across 41 benchmark runs. Zero LLM calls. Sub-millisecond latency.

Cost calculator

Your agent re-sends its full tool list on every call. That adds up.

Every time your agent talks to the AI, it re-explains what tools it can use — search, read file, call API, whatever. That description is hundreds of words, sent thousands of times a day. Move the sliders to see what it costs.

How many tools does your agent have? 10

1 tool64 tools

How many times does it call the AI per day? 1,000

100 calls/day100,000 calls/day

Tokens saved / day

—

Saved / month

—

Saved / year

—

Each tool definition is ~120 words sent to the AI. Pricing based on Claude Sonnet 4 (~$3 per million words). First call: 69% smaller. Every call after: 97% smaller. Average across a session: ~93%. How does this work? →

Pricing

Start free. Pay when you need more.

Free

forever

~15–20% shorter — runs entirely in your browser. No server, no account, fully private.