sebbi.pro
Token saver

Stop paying for the same answer twice.

A single file you run on your own machine. It sits between your application and your model provider, and it never lets you buy an answer you have already bought. Your integration is one line.

wasbase_url = "https://api.anthropic.com"
nowbase_url = "http://127.0.0.1:8788"
Download sebbi_tokensaver.py
Free for 7 days, then 50p per machine per month. Runs with no account at all if you just want the saving.

Four ways it takes money off your bill

None of these involve calling another model. Governance that costs tokens to run is the thing this replaces.

Free

An identical request is answered from your own machine

Same model, same messages, same settings — it never reaches your provider. The answer is stored on your disk, not ours, and comes back byte for byte.

Stopped

A runaway agent is caught in microseconds

The same call going round eight times in two minutes is a loop, not a workload. It gets refused locally before the money goes. Four times if you have told it no human is watching.

Stopped

A spend ceiling that actually holds

Set a token ceiling. When it is reached, requests are refused before they reach the model. A single call that could cost more than the budget left is held before it spends, not reported after.

Named

It tells you what is wasteful in each request

Thirty turns re-sent on every call. Twelve tool definitions nothing has ever used. Temperature above zero on a request that did not need to vary, which is the only reason its answer cannot be reused. This is the part that changes what you send tomorrow.

What leaves your building

Paste a real request on the left. The right side is computed in your browser, right now, and is exactly what the file would send us. Nothing is uploaded by this page.

Your requeststays on your machine
The digestall we ever see

A SHA-256 fingerprint and five counts. There is no way to read a prompt back out of a hash — change one character on the left and watch the whole fingerprint change, with nothing about the change visible in it. Run it with --offline and even the digest stays home.

Running it

Python 3.8 or newer. No dependencies to install — it uses only what ships with Python.

  1. Start it

    python3 sebbi_tokensaver.py --key YOUR_KEY --provider anthropic

    No key yet? Use --offline. It saves you exactly the same money; you just do not get receipts.

  2. Change the one line

    base_url = "http://127.0.0.1:8788"   # that is the whole integration
  3. Watch it work

    http://127.0.0.1:8788/saver

    One number: tokens you did not buy, taken from your provider's own reported counts. Never an estimate, never a percentage.

Getting a key

The file is free and always will be. A key is what turns your saving into a receipt: every decision sealed into a chain, a ledger you can hand to your finance team, and spend ceilings that hold across every machine you run.

Free for 7 days — the full thing, no card. After that it is 50p per machine per month, billed through Stripe, counted on the machines that actually used your key rather than a number you typed. Stop whenever you like; the file carries on saving you money without us.

What it does not do

Published here rather than discovered later.

AI news
The GuardianCalifornia sues OpenAI over safety concerns after rogue agent hacking10mThe GuardianAlbanese signs on to Trump’s AI supply chain pact before leaders’ meeting at Quad summit50mThe VergeCalifornia sues OpenAI over rogue AI hacking incidents1hTechCrunchCalifornia state AG sues OpenAI over rogue agent incidents1hThe VergeApple is testing daily usage limits for more complex Siri AI interactions2hThe VergeOpenAI wants to stop deepfakes from overwhelming child abuse tiplines2hThe DecoderOpenAI's GPT-6 Astra tackles the Standard Model's toughest math problem on a lattice2hBBC NewsTrump backs return of ex-Nvidia boss after 'Bring Jensen Back' campaign launched2hTechCrunchTrump says ex-Nvidia CEO Jensen Huang’s affair with an alleged Chinese spy did not threaten national security2hArs TechnicaUS moves to cut China’s remote access to AI chips2hThe DecoderMoonshot AI pivots Kimi models from coding to science as benchmark gains flatten2hThe VergeNvidia’s board is reportedly ‘not considering’ hiring Jensen Huang back3hThe GuardianTrump backs calls for reinstatement of Jensen Huang as Nvidia CEO3hThe GuardianAnthropic recognises the risks to humanity – but is its governance really up to the job? | Nils Pratley3hTechCrunchSam Altman on Nvidia without Jensen Huang: the industry will be ‘fine’3hTechCrunchFBI says Artex AI open-source hacking tool used to breach networks in ‘under three hours’4hTechCrunchJordan Belfort’s new AI voice startup, Tone, wants to sell like the ‘Wolf of Wall Street’4hThe DecoderAlibaba's Qwen4-Max takes on Opus 5.5, GPT-6, and Kimi K3.1 Pro at much lower prices4hThe DecoderClaude now builds interactive websites inside canvas graphics, reviving the Flash era with real HTML5hThe GuardianAI models ‘weak and fallible’ on climate science, UN-backed study reports5hMIT Technology ReviewBuilding trustworthy AI by fencing it in, not predicting it5hThe DecoderAnthropic says the key to agent memory is knowing what to forget, not how much to store6hArs TechnicaJensen Huang allies build a "war room" to get him back at Nvidia7hThe VergeThe Nvidia scandal isn’t slowing down8hThe GuardianCalifornia sues OpenAI over safety concerns after rogue agent hacking10mThe GuardianAlbanese signs on to Trump’s AI supply chain pact before leaders’ meeting at Quad summit50mThe VergeCalifornia sues OpenAI over rogue AI hacking incidents1hTechCrunchCalifornia state AG sues OpenAI over rogue agent incidents1hThe VergeApple is testing daily usage limits for more complex Siri AI interactions2hThe VergeOpenAI wants to stop deepfakes from overwhelming child abuse tiplines2hThe DecoderOpenAI's GPT-6 Astra tackles the Standard Model's toughest math problem on a lattice2hBBC NewsTrump backs return of ex-Nvidia boss after 'Bring Jensen Back' campaign launched2hTechCrunchTrump says ex-Nvidia CEO Jensen Huang’s affair with an alleged Chinese spy did not threaten national security2hArs TechnicaUS moves to cut China’s remote access to AI chips2hThe DecoderMoonshot AI pivots Kimi models from coding to science as benchmark gains flatten2hThe VergeNvidia’s board is reportedly ‘not considering’ hiring Jensen Huang back3hThe GuardianTrump backs calls for reinstatement of Jensen Huang as Nvidia CEO3hThe GuardianAnthropic recognises the risks to humanity – but is its governance really up to the job? | Nils Pratley3hTechCrunchSam Altman on Nvidia without Jensen Huang: the industry will be ‘fine’3hTechCrunchFBI says Artex AI open-source hacking tool used to breach networks in ‘under three hours’4hTechCrunchJordan Belfort’s new AI voice startup, Tone, wants to sell like the ‘Wolf of Wall Street’4hThe DecoderAlibaba's Qwen4-Max takes on Opus 5.5, GPT-6, and Kimi K3.1 Pro at much lower prices4hThe DecoderClaude now builds interactive websites inside canvas graphics, reviving the Flash era with real HTML5hThe GuardianAI models ‘weak and fallible’ on climate science, UN-backed study reports5hMIT Technology ReviewBuilding trustworthy AI by fencing it in, not predicting it5hThe DecoderAnthropic says the key to agent memory is knowing what to forget, not how much to store6hArs TechnicaJensen Huang allies build a "war room" to get him back at Nvidia7hThe VergeThe Nvidia scandal isn’t slowing down8h