| What it protects | Whatever your code passes to it. It ships detectors for items such as credit card numbers and US Social Security numbers, plus name detection through an NLP model. You can add your own detectors. | Names, identifiers, addresses, secrets and anything on your own "always protect" list, in text you type or paste. |
|---|
| Where processing happens | Wherever you run it: your own servers, containers or data pipelines. | On your device. |
|---|
| Where original values are stored | Wherever your code keeps them. With the encrypt operator, the original comes back from the encrypted value and your key. | In a swap table that stays on your device. |
|---|
| Restore in replies | Possible, if you build it. Its deanonymizers revert operations such as encryption. Wiring that into AI replies is your work. | Yes. Real values return on your device in supported replies. |
|---|
| How you use it | A Python library, or REST services in Docker. You call the analyzer, then the anonymizer, inside your own app. | A browser extension for ChatGPT, Claude, Gemini, Copilot and Perplexity, a web workspace, or the SDK and local proxy in your own code. |
|---|
| Deployment | Self-hosted, from Python or PySpark through Docker to Kubernetes. Docker images now publish to the GitHub Container Registry. | Extension version 0.1.1 installs manually in Chrome with "Load unpacked". Workspace and billing run on tightlipai.com during the move to the new name. |
|---|
| Pricing | Free. MIT license. You pay for your own hosting and engineering time. | USD 0.05 per credit. Top-ups from USD 100. USD 20 free credit on sign-up (to be reconfirmed before launch). One credit covers each started block of 8,000 tokens. |
|---|