Local by default
Text extraction, entity detection, and rewriting all run on your machine using a bundled local model.
Anonymize on your machine before the LLM sees it
Piixie
Scrub private data from documents, tables, and notes before you paste, upload, or automate anything with an LLM.
- Redact, replace, or synthesize
- Runs on your machine
- Desktop, server, and CLI workflows
- Bring your own LLM when needed
Privacy before the prompt
Documents stay on your machine
Works with any LLM tool
For solo use and teams
Anonymize before the prompt leaves your desk.
You use LLMs to get work done. You should not have to paste raw PII into someone else's service to do it. Piixie strips private data on your machine so you send only what you intend.
Flexible outputs
Strict redaction, stable replacement tokens, or synthetic values that keep the same format.
Automation ready
Run Piixie interactively, from the CLI, or as an internal server for batch jobs and shared team access.
Three modes, depending on what you need.
Some workflows need the strongest redaction. Others need the document to stay readable. You pick.
- Redaction removes sensitive values entirely.
- Replacement keeps stable placeholders for review and tracing.
- Synthetic mode keeps document structure for demos, tests, and prompts.
Product screenshots for the work teams actually do.
Desktop review, synthetic replacement tables, and CLI automation all use the same local pipeline.
Redaction
Replace names, emails, phone numbers, and other PII with blacked-out blocks.
Synthetic
Swap in fake names and addresses that keep the document readable.
CLI and server
Batch documents from scripts or run a shared server for the team.
Synthetic data that reads like the original.
The local model spots private entities and picks a generator for each one. Piixie then produces fake values on your machine using Faker-backed profiles.
- Names, addresses, companies, phones, accounts, cards, IDs, URLs, IPs, and dates all use typed Faker generators.
- Repeated mentions of the same person, company, or address reuse one cached fake profile so the document stays coherent.
- Values with specific formats keep their shape: ID patterns, birthdate separators, and email local-part style.
Faker-backed
Fake values come from a bundled local Faker runtime.
Consistent
One detected entity gets one fake identity across names, addresses, and linked values where possible.
Shape-aware
IDs and dates keep their formatting so tests, demos, and prompts still look right.
Local-first architecture, optional external LLMs.
Piixie keeps as much as possible on the device. When teams use external models, they send an already-anonymized copy.
How Piixie fits into an LLM workflow.
The same pipeline runs on the desktop, at the command line, and on a shared server.
flowchart LR
A[Raw document] --> B[Extract locally]
B --> C[Detect PII]
C --> D{Choose mode}
D --> E[Redact]
D --> F[Replace]
D --> G[Synthesize]
E --> H[Safe copy]
F --> H
G --> H
H --> I[Prompt, share, archive, or test]
Pricing based on how you run it.
Start free. Move to Pro when you anonymize documents daily. Run a shared server when the whole team needs it.
Free
For occasional local anonymization.
$0/ month
- 5 local anonymizations per day
- Desktop app access
- Redaction, replacement, and synthetic modes
Pro
For people anonymizing documents every workday.
$10/ month
- 20 daily local anonymizations
- Short-burst local processing
- Preferred local-first workflows
Scale-up
For teams that need a shared internal service.
Talkto us
- Central Piixie server setup
- Multiple users and higher daily volume
- Reduced individual workstation load
Enterprise
For organizations that need SSO and audit controls.
Custompricing
- Centralized authentication
- Single sign-on and security reviews
- Custom features and support
Docs, changelog, and guides are live.
Setup, supported file types, automation, server mode, and product updates.