Microsoft Presidio alternative: self-hosted Presidio or a hosted PII and secrets redactor?
Short answer: Presidio, now a community-owned project called “Presidio by Data Privacy Stack” (the old microsoft/presidio repository redirects to it), is free, open-source software that you install and run yourself. It finds PII in text and images with NER models, patterns and checksums, supports multiple languages and lets you add your own recognizers. The Fairquery PII redactor is a hosted Apify Actor with a fixed set of pattern-based detectors for common PII plus 18 named secret types such as API keys, private keys and JWTs. It costs $1.50 per 1,000 documents plus $0.20 per 1,000 findings on Apify’s Free plan, with nothing to host. If you can run a Python service, need many entity types or languages, image redaction or data that never leaves your servers, Presidio fits better. If you want secrets detection and PII redaction for LLM prompts without running anything, the Fairquery redactor may fit better.
Price side by side, same unit
Here one result is one document: one prompt, ticket, file or URL processed. Fairquery figures are the Apify pay-per-event prices from the Actor's live Apify pricing, checked 2026-10-30, on the Apify Free plan unless a row says otherwise. Presidio figures come from its GitHub repository (MIT license); the cost of hosting it depends on your own setup, so no figure is given for it. Apify plan figures from Apify pricing, checked 2026-10-30.
| Presidio (self-hosted) | Fairquery PII redactor | |
|---|---|---|
| Price per 1,000 documents | $0.00 for the software (MIT license). Your own hosting and upkeep are extra; their cost depends on your setup and is not published | $1.50 per 1,000 documents, plus $0.20 per 1,000 findings (Free-plan prices)* |
| Cost for 1,000 results | $0.00 for the software, plus your hosting | $1.50, plus $0.20 per 1,000 findings* |
| Cost for 10,000 results | $0.00 for the software, plus your hosting | $15.00, plus $0.20 per 1,000 findings* |
| Lower prices at volume | Not applicable | $1.35 per 1,000 documents on Apify Starter, $1.20 on Scale, $0.99 on Business; findings stay $0.20 per 1,000 |
| Free tier | The whole framework is free to install and run | No Fairquery free tier. Apify Free: $5 a month of platform credit, about 3,333 documents with no findings |
* Plus a start fee of $0.05 per 1,000 runs (one start event per GB of run memory, at least one per run). Error rows are not charged.
Result: Presidio’s software is free, so if you already run Python services and can host it, it costs less than the Fairquery redactor at any volume. The Fairquery redactor costs less only when the time and servers to deploy and maintain Presidio cost you more than the usage fees, for example for small or irregular workloads.
Who each tool fits best
The Presidio points below come from its documentation; the Fairquery points from the redactor Store page.
Presidio fits better if you…
- can run a Python service, a Docker container or a Spark job yourself;
- need data to stay on your own infrastructure;
- need many entity types, country-specific IDs or languages other than English, or want to add your own recognizers and models;
- need to redact PII in images, or work on structured data with its structured module;
- want free software with no usage fees.
The Fairquery redactor fits better if you…
- want to call an API or an MCP tool and have nothing to install or host;
- need leaked API keys, tokens, private keys and JWTs caught alongside PII, without writing custom recognizers;
- want reversible placeholders such as [EMAIL_1] and a restore mode for the LLM’s answer;
- process text, HTML or text-based PDFs in English;
- prefer paying per document, so a small or irregular workload costs cents.
Presidio and the Fairquery redactor compared
Facts about Presidio come from its own documentation and repository, checked on 2026-10-30; each row names its sources, and the full list of sources with check dates is at the end of this page.
| Presidio | Fairquery PII redactor | Sources | |
|---|---|---|---|
| What it is | Open-source framework for finding and anonymizing PII in text and images (analyzer, anonymizer, image redactor and structured-data modules); now a community-owned project, “Presidio by Data Privacy Stack” | Hosted Apify Actor that detects and redacts PII and secrets in text, HTML and PDF before it reaches an LLM | Presidio home; redactor Store page |
| How you run it | Python 3.11 to 3.14 packages with pip and an NLP engine (spaCy by default, or transformers or Stanza), Docker, or an HTTP service; Python or PySpark workloads up to Kubernetes | Call it through the Apify Console, API or Apify MCP; nothing to install | Presidio installation; Presidio home; redactor Store page |
| How it detects | Named Entity Recognition, regular expressions, rule-based logic and checksums with context words; options for connecting external PII detection models | Patterns with validation (Luhn and issuer prefix for cards, mod-97 for IBANs, libphonenumber for phones, SSN rules); opt-in English names via open-source NER; Gitleaks-style secret rules and an entropy threshold | Presidio home; redactor Store page |
| PII types | Global types such as CREDIT_CARD, CRYPTO, DATE_TIME, EMAIL_ADDRESS, IBAN_CODE, IP_ADDRESS, MAC_ADDRESS, NRP, LOCATION, PERSON, PHONE_NUMBER, MEDICAL_LICENSE and URL, plus country-specific recognizers (for example USA, UK, Spain, Italy, Poland, Singapore, Australia, India, Finland, Korea, Thailand, Nigeria, Germany, Sweden and Canada); custom recognizers can be added | EMAIL, PHONE, SSN, CREDIT_CARD, IBAN, UK_NINO, IP_ADDRESS; opt-in PERSON, LOCATION, STREET_ADDRESS | Presidio supported entities; redactor Store page |
| Secrets (API keys, tokens) | No API-key or token entity on the supported entities page; you can add custom recognizers | 18 named types, such as AWS, GitHub, GitLab, Slack, Stripe, Google, OpenAI and Anthropic keys, private keys and JWTs, plus generic secrets (api_key=, token:, password=) and opt-in high-entropy strings | Presidio supported entities; redactor Store page |
| Languages | Multiple languages; English recognizers and models by default, other languages need an NLP model and recognizers | Patterns work on any text; name detection is English only | Presidio multi-language support; redactor Store page |
| Redaction and restore | Operators such as replace, mask, redact and encrypt; deanonymizers revert an operation, for example decrypting | Placeholder (default), mask or salted hash; placeholder map plus restore mode to put original values back | Presidio anonymizer; redactor Store page |
| Images and scanned PDFs | Image redactor using OCR | No OCR: scanned PDFs can slip through | Presidio home; redactor Store page |
| Where the data goes | Wherever you deploy it, so data can stay on your own infrastructure | Processed in your Apify account; designed not to log document content and, by default, tries to delete the run’s input when it finishes | Presidio installation; redactor Store page |
| LLM pipeline recipes | Recipes for masking LLM calls with the LiteLLM proxy and the AISIX AI Gateway | Apify API or Apify MCP from an agent; FAQ and examples for redact, send output_text, restore | Presidio home; redactor Store page |
What it costs: three example workloads
Fairquery figures use Free-plan prices unless the row names another Apify plan, and leave out the start fee (see the footnote above). Above $5 of usage a month you need an Apify paid plan (Starter: $19/month, includes $19 of usage). Presidio rows show only the software price, because hosting depends on your setup.
| Workload | Presidio (self-hosted) | Fairquery PII redactor |
|---|---|---|
| 1,000 support tickets, 2,000 findings | $0.00 for the software, plus your hosting | $1.90 (1,000 at $1.50 per 1,000 plus 2,000 at $0.20 per 1,000); inside Apify Free's $5 |
| 10,000 prompts, 5,000 findings | $0.00 for the software, plus your hosting | $16.00 ($15.00 plus $1.00) on Free-plan prices; above $5 of usage a month you need an Apify paid plan |
| 100,000 documents, 50,000 findings, Apify Business plan | $0.00 for the software, plus your hosting | $109.00 ($99.00 at $0.99 per 1,000 plus $10.00 at $0.20 per 1,000) |
Neither tool finds everything. Presidio’s documentation warns that there is no guarantee it will find all sensitive information, and the Fairquery redactor’s Store page says unusual formats, non-English names and scanned PDFs can slip through. Use either one as one layer of data protection, not as a compliance certification.
How to try the Fairquery redactor
- Open the Actor on the Apify Store and paste texts, documents or public document URLs.
- Pick
redactmode and a redaction style (placeholder,maskorhash); turn on names and addresses if you need them. - Send
output_textto your model. To restore values, turn onreturnPlaceholderMapand run the answer back inrestoremode.
A longer walkthrough is in the guide how to redact PII and secrets before sending data to an LLM.
FAQ
Is the Fairquery redactor a drop-in replacement for Presidio?
No. Presidio is an open-source Python framework that you install and run yourself, with NER models, many entity types, country-specific recognizers, image redaction and custom recognizers. The Fairquery PII redactor is a hosted Apify Actor with a fixed set of pattern-based detectors for common PII and 18 named secret types. It replaces Presidio only if that fixed set covers what you need and you would rather not run the service yourself.
Presidio is free. Why pay for a redactor?
Presidio itself costs nothing: it is MIT-licensed open source. You still run it, so you pay for your own servers and the time to install models, deploy and maintain it. The Fairquery redactor costs $1.50 per 1,000 documents plus $0.20 per 1,000 findings on Apify's Free plan, with nothing to host. If you already run Python services, Presidio is usually the cheaper choice.
Does Presidio detect API keys and other secrets?
Presidio's supported entities page, checked on 2026-10-30, lists PII types such as credit cards, emails, IBANs, IP addresses, names, locations and phone numbers, plus country-specific IDs, and no API-key or token entity. Presidio lets you add custom recognizers, so you can write your own. The Fairquery redactor detects 18 named secret types, such as AWS, GitHub, Stripe, OpenAI and Anthropic keys, private keys and JWTs, plus generic secrets like password=...
Which one is better for languages other than English?
Presidio. Its documentation says it supports PII detection in multiple languages; the default configuration has recognizers and models for English, and you add an NLP model and recognizers for other languages. The Fairquery redactor's name detection is English only, and its Store page says non-English names can slip through.
Can either one put the original values back after the LLM answers?
Yes. Presidio's anonymizer has deanonymizers that revert an operation, for example decrypting an encrypted value. The Fairquery redactor returns a placeholder map when you turn on returnPlaceholderMap, and restore mode puts the original values back into the model's answer.
Is Fairquery affiliated with Microsoft or Presidio?
No. Microsoft and Presidio are trademarks of their owners. Fairquery is not affiliated with, endorsed by or sponsored by Microsoft, Presidio or Data Privacy Stack. The names are used only to compare the two tools.
Sources
- Presidio documentation home: description, main features (NER, regex, rule-based and checksum recognizers in multiple languages; external models; Python, PySpark, Docker, Kubernetes; image redaction), modules, the warning that detection has no guarantee, community-owned transition, LiteLLM and AISIX recipes. Checked 2026-10-30.
- Presidio supported entities: global and country-specific entity types and detection methods; custom recognizers. Checked 2026-10-30.
- Presidio installation: Python 3.11 to 3.14; pip with spaCy, transformers or Stanza; Docker; image redactor needs OCR. Checked 2026-10-30.
- Presidio anonymizer: operators (replace, mask, redact, encrypt) and deanonymizers. Checked 2026-10-30.
- Presidio multi-language support: English by default; extra languages need an NLP model and recognizers. Checked 2026-10-30.
- Presidio GitHub repository: MIT license; the old microsoft/presidio address redirects here. Checked 2026-10-30.
- Apify pricing: Free plan with $5 to spend; Starter $19/month with $19 to spend. Checked 2026-10-30.
- Fairquery PII redactor on the Apify Store: pay-per-event prices, detected types, redaction styles, input, data handling and limits; prices also read from the public Apify API. Checked 2026-10-30.
Presidio details checked on 2026-10-30 from its public documentation and repository; they may change. Microsoft and Presidio are trademarks of their owners. Fairquery is not affiliated with, endorsed by or sponsored by Microsoft, Presidio or Data Privacy Stack; the names are used only to identify the product being compared. This page is not legal advice. Corrections are welcome at info@getfairquery.com.
See also all Fairquery tools, all guides and tools for AI agents.