Pragma Redact Keeps the Redaction in Your Browser

Field notes

The tool separates three input paths—PDF, plain text, and image OCR—and describes a browser-side sequence: download the detection model, identify sensitive spans, then review or copy the result. Its examples include scrubbing chatbot prompts, support logs, screenshots, README files, API keys, and tokens. The site says the model is retained in browser storage after its first download and that later runs can use available graphics acceleration; older mobile hardware may be a limitation.