# Redact PDF AI

Aiolum story: https://aiolum.com/story/redact-pdf-ai/ · added 2026-11-03 · updated 2026-11-03

**In short.** Redact PDF AI (AI redaction of personal data in PDF files, Switzerland): US$1.9K in verified revenue since launch, October 2026. Evidence grade A: disclosed by a company or a filing, or verified by a payment processor. Team: 1 founder; headcount not published. Launched 30 November 2025; still running, revenue falling.

AI redaction of personal data in PDF files · Switzerland · AI-native · solo founder · still running, revenue falling · group: Documents, data and slides

- Outcome: US$1.9K in verified revenue since launch, October 2026 (evidence grade of the headline number: A; headline in US dollars: 1919)
- Launch: 30 November 2025; months from launch to the outcome: 11
- Team: 1 founder; headcount not published
- Outside funding: No outside funding found in any source
- Main factor: Niche
- Meters, 1 to 4: evidence strength 3, copyable pattern 2, capital needed 1
- Tags: AI-native, Solo founder, Bootstrapped, Running, B2B, Privacy

Links:

- Visit redact-pdf.ai: https://www.redact-pdf.ai/
- Verified revenue on TrustMRR: https://trustmrr.com/startup/redact-pdf-ai

**The product.** A web app and REST API that finds personal data in PDF files and images (names, emails, phone numbers, addresses, organizations, dates, IBANs, card numbers) and removes it for good. The user reviews the AI's masks in an editor, adds or removes some, and downloads a file whose pages are turned into images, so no hidden text stays under the black boxes.

**Why it worked.** Accountants, property managers, real estate agents and medical offices must share documents that name other people. Drawing black boxes in a PDF editor is slow and often leaves the text in the file. This tool does the search with AI and sells it per page, with a first document free and plans from US$50 a month.

Evidence grades: A = disclosed by a public company, a filing or a payment processor, or verified by one; B = two or more independent reports agree, or a journalist confirmed it; C = a founder claim that no one else verified; D = reported, not verified: one founder claim or a third-party estimate with no second source; calc = computed on this page from graded facts.

## Summary

Redact PDF AI is a redaction tool. Redaction means removing confidential text from a document before it is shared, so that nobody can read or recover it. The product looks for PII (personally identifiable information: data that points to one person, such as a name, an address or a bank account number) with OCR, optical character recognition, which reads the text in scanned pages, and with Microsoft Azure's AI services for PII detection. It then rasterizes each page, which means it turns the page into a picture, and burns the black masks into the pixels. A REST API and an MCP server let other programs and AI agents call it.

The founder is Damien Buchs, named as data controller in the privacy text of the site, with an address in Morges, Switzerland, and signing the blog as "Founder of Redact PDF AI". TrustMRR, a site that reads revenue straight from a company's payment account, gives the founding date as 30 November 2025 and shows the first revenue in January 2026. Its Stripe connection shows US$1,919 in total revenue up to its last sync on 11 October 2026, a best month of US$553 in May 2026, and US$163 in October 2026 up to that day. At the last sync it reported MRR, monthly recurring revenue from subscriptions, of US$350 from 7 active subscriptions.

These revenue numbers are grade A, because they come from the payment processor, but they stopped on 11 October 2026, when the Stripe API key that TrustMRR used expired. It is a small business: the money in total is less than one month of income for most stories on this site. The page shows it as an example of a narrow, regulated-sounding task sold by one person, and it compares it with earlier companies that sold privacy or document handling and failed.

## Pricing

Redact PDF AI prices read from the pricing section of the company's home page on 3 November 2026 (grade C)

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Prepaid credits, first pack | 5 one-time |  | 50 pages, use within 30 days |
| Prepaid credits | 10 one-time |  | 50 pages (US$0.10 a page), use within 30 days |
| Starter |  | 50 | 1,000 pages a month (US$0.05 a page), up to 1,000 pages a document |
| Business |  | 250 | 6,000 pages a month (US$0.04 a page), up to 3 seats, phone support |
| Enterprise | contact sales |  | uncapped pages, SSO / SAML, unlimited seats, custom pipelines |

The first document, up to 5 pages, is free with no credit card. Credit packs are one-time payments valid for 30 days; the first pack costs US$5 instead of US$10. The page shows monthly subscription prices only, with no yearly option. Re-running the analysis on a document is free.

## How much revenue does Redact PDF AI make? The numbers, graded

A one-founder redaction tool with US$1,919 of Stripe-verified revenue in about ten months, MRR of US$350 at the last sync on 11 October 2026, and no data after that day

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Total verified revenue, to 11 October 2026 | US$1,919 | A | TrustMRR, Stripe API source. The "All time" and "Last 12 months" rows both show US$1,919. Revenue last synced 2026-10-11T14:32:21Z. | [1] |
| Stripe connection | API key expired on 11 October 2026 | A | TrustMRR writes: "Stripe API key expired at 2026-10-11T20:32:21.022Z". The daily table shows US$0.00 for every day after 8 October, which means no data, not proof of no sales. The live profile read on 3 November 2026 gives the same figures as the copy downloaded on 31 October 2026. | [1] |
| Best month | US$553, May 2026 | A | TrustMRR monthly revenue timeline. | [1] |
| Monthly revenue, January to May 2026 | US$65.18, US$10.89, US$85.78, US$153, US$553 | A | TrustMRR monthly revenue timeline. The first month with revenue is January 2026. | [1] |
| Monthly revenue, June to October 2026 | US$240, US$256, US$195, US$198, US$163 | A | TrustMRR monthly revenue timeline. October covers only the days up to the last sync on 11 October; the daily table shows US$56.25 on 6 October and US$50.00 on 8 October. | [1] |
| MRR and subscriptions at the last sync | US$350 a month, 7 active subscriptions | A | TrustMRR metric snapshots: "Current MRR: $350", "Current active subscriptions: 7", MRR growth +133.3% in the last 30 days. Revenue in the last 30 days before the sync: US$355. | [1] |
| MRR as a yearly run rate | US$4,200 a year | calc | Calculated: US$350 times 12. Monthly revenue from June to September 2026 averaged US$222.25, below this rate, so the subscriptions had grown shortly before the sync. | [1] |
| Seven subscriptions and US$350 of MRR | matches 7 Starter plans at US$50 | calc | Calculated: 7 times US$50 is US$350. TrustMRR does not publish the plan mix, so this is only one way to reach the figure. | [1, 2] |
| Share of all revenue earned in May 2026 | about 29% | calc | Calculated: US$553 divided by US$1,919. | [1] |
| Change from April to June 2026 against July to September 2026 | US$946 down to US$649, about 31% lower | calc | Calculated from the monthly timeline. This page therefore marks the trend as falling. | [1] |
| Founding date | 30 November 2025 | C | Founder-entered field on TrustMRR. The oldest posts on the company blog are dated 6 December 2025. | [1, 4] |
| Founder | Damien Buchs, Morges, Switzerland | C | The privacy text on the site says: "Data controller: Damien Buchs", with an address in Morges, Switzerland (last updated 24 April 2026). The blog signs posts "Damien B., Founder of Redact PDF AI". TrustMRR lists one founder, Damien, country CH. | [3, 5, 1] |
| Team | 1 founder named; headcount not published | C | No source names any other person or gives a headcount. The Business plan offers phone support. | [1, 2] |
| Outside funding | none found | C | No source mentions investors or a funding round. | [1, 2] |
| AI services and hosting | Microsoft Azure in Frankfurt and Switzerland North | C | The security page says content logging is disabled "on the Azure AI services used for OCR and PII detection". Subprocessors named in the privacy text: Microsoft Azure, Clerk, Stripe, Mixpanel, Sentry, Google Ads, Supabase and Loops. | [3] |
| Compliance status | the company itself is not audited | C | Home page FAQ: "Redact PDF AI itself is not independently audited for these frameworks". The SOC 2, ISO 27001, ISO 27017 and ISO 27018 certificates and the HIPAA eligibility belong to Microsoft Azure. | [2, 3] |
| Usage claims on the home page | "> 10,000" document pages redacted, "99.9%" accurate | C | Company claims, not verified. The page count does not conflict with the revenue: at US$0.04 to US$0.10 a page, US$1,919 pays for more than 10,000 pages. No test or source is given for the accuracy figure. | [2] |
| Developer access | REST API, keyless demo endpoint, open-source MCP server | C | Blog post of 22 August 2026: three calls (upload, poll, download) with an idempotency key; the MCP package redact-pdf-mcp is on npm and in the official MCP registry. | [5] |
| Languages | 100+ document languages, mixed-language files | C | Home page. | [2] |
| Time from founding to the last verified figure | 11 months | calc | Calculated: November 2025 to October 2026. | [1] |

How to read this table: a grade A value comes from the Verification Sources section of the TrustMRR profile, which reads the company's Stripe account through an API key. A grade C value comes from the company's own website or from text that the founder wrote on TrustMRR. A grade X value was computed on this page from rows above it.

## When did Redact PDF AI launch, and how did it grow? Timeline

From a founding date in late November 2025 to a best month in May 2026 and a stop in verified data on 11 October 2026. Each entry carries its grade.

- **30 November 2025** (grade C [1]): Founding date entered on the TrustMRR profile.
- **6 December 2025** (grade C [4]): The first posts appear on the company blog, followed by more on 10, 14 and 20 December.
- **January 2026** (grade A [1]): First verified revenue: US$65.18.
- **April 2026** (grade A [1, 3]): Revenue reaches US$153 in the month. The privacy text naming Damien Buchs as data controller is last updated on 24 April.
- **May 2026** (grade A [1]): Best month: US$553.
- **June to September 2026** (grade A [1]): Revenue between US$195 and US$256 a month.
- **22 August 2026** (grade C [5]): A blog post by the founder describes the REST API for AI agents and announces an open-source MCP server.
- **11 October 2026** (grade A [1]): Last Stripe sync: US$1,919 in total, MRR US$350 from 7 subscriptions. The Stripe API key used by TrustMRR expires the same day.
- **12 October 2026** (grade C [4]): Four blog posts on removing names from documents before pasting them into ChatGPT, Claude or Gemini.
- **3 November 2026** (grade C [2]): The site is live with prices from US$5 for a credit pack to US$250 a month.

## Why did Redact PDF AI work?

### A narrow job with a clear buyer

The site speaks to four groups that share documents naming other people: accounting and fiduciary offices, real estate, property management and healthcare. Each has its own landing page and lists of document types (tenant applications, lease agreements, lab results, powers of attorney).

### Rent the AI, sell the workflow

The detection of personal data and the OCR run on Microsoft Azure's AI services. What the founder built is the work around it: page-by-page progress, an editor to fix the masks, excluded terms, batch upload, retention settings and irreversible output.

### Per-page prices with a free first document

A visitor can redact page 1 of a file without an account, and the first document of up to 5 pages is free. After that, a first credit pack costs US$5 and subscriptions start at US$50 a month for 1,000 pages.

### Many pages for search, and an API for agents

The site has pages for each kind of data and document, comparison pages against Adobe Acrobat and other tools, and a blog. In August 2026 it added a keyless demo endpoint and an MCP server so that AI agents can call it. No source says which of these brought the paying customers.

## How it was built

These details come from the company's own home page, security page, privacy text and blog. They are grade C.

- Microsoft Azure for hosting, Blob storage, OCR and PII detection, in the Frankfurt and Switzerland North regions, with keys in Azure Key Vault and a managed PostgreSQL database.
- Clerk for login and accounts, Stripe for payments, Mixpanel for analytics after consent, Sentry for error monitoring, Supabase and Loops for some demo flows that send download links by email.
- Pages are rasterized and the masks are burned into the pixels, so the output has no hidden text layer and no recoverable metadata.
- Files are deleted automatically after 3, 7 or 14 days of inactivity (14 by default), or at once on request.
- A REST API with asynchronous jobs, an idempotency header for safe retries, an OpenAPI file, llms.txt and Markdown docs, and an open-source MCP server published on npm.
- The site is in English, French, Spanish, German, Japanese and Portuguese.

Sources: [2, 3, 5]

## Who tried this before Redact PDF AI, and why did they fail?

Five companies in our database of failed startups sold privacy, sensitive records or AI for document-heavy professions, and failed. The values below are from that database and are not independently verified.

The failed startups come from our database of dead startups, retold here in our own words; Its entries are AI-assisted summaries of public sources, so they carry no evidence grade.

### Graphite Docs (2018 to 2020; funding US$400K; cause: No Market Need)

- Promise: A private, encrypted alternative to Google Docs.
- Why it died: Users preferred free and fast tools over secure and more expensive ones.
- Record: 

### Skiff (2020 to 2024; funding US$14.2M; cause: No Market Need)

- Promise: End-to-end encrypted email, documents and storage as a private alternative to Google Workspace.
- Why it died: Privacy features were not enough to pull mainstream users away from free tools with network effects.
- Record: 

### ROSS Intelligence (2015 to 2018; funding US$13M; cause: Competition)

- Promise: AI legal research that would be faster than the old legal databases.
- Why it died: Lawyers stayed with the databases they already used, sold by much larger companies.
- Record: 

### PatientBank (2016 to 2019; funding US$250K; cause: No Market Need)

- Promise: Patients request and share their own medical records online.
- Why it died: Large healthcare organizations and their rules slowed sales, and few buyers were ready for it.
- Record: 

### Lendsnap (2016 to 2019; funding US$500K; cause: Competition)

- Promise: Collect a mortgage borrower's financial documents automatically.
- Why it died: A single document feature could not compete with all-in-one mortgage platforms such as Blend.
- Record: 

## What changed

The predecessors show two ways that privacy and document products die. Graphite Docs, Skiff and PatientBank asked people to change a habit for privacy, and most people did not. ROSS Intelligence and Lendsnap sold one function to professions that already had large suppliers, and the suppliers won.

Redact PDF AI asks for no change of habit: a person who must send a file that names other people already needs to black out those names, and the tool does it on one file at a time, paid per page. It also did not have to build its own AI model, because Azure sells PII detection and OCR as services. That keeps the cost low enough for one person to run it. The risk from the predecessors remains: Adobe Acrobat and macOS Preview already have redaction tools, and the founder's own blog explains how to use them.

## Can one person copy Redact PDF AI now?

**Verdict: Yes, with modest expectations.**

One person built a working product on rented AI services, with an API, an MCP server, six languages and many search pages, in under a year. Stripe data shows it earned US$1,919 in that time, so this is a model for a small side income, not yet for a living. The method can be repeated in any field where a document must be cleaned before it is shared.

1. Pick one document task that a regulation or a client forces on people, such as removing names before sharing.
2. Use a cloud provider's AI service for the hard part, and build the review editor and the output around it.
3. Let visitors try one page without an account, then charge per page or per month.
4. Add an API and an MCP server so that other programs and AI agents can buy the same service.
5. Connect the payment account to a public revenue site, so that buyers and readers can check the numbers.

**What is still open.** Other cleaning steps before a document goes into an AI model: removing names consistently from legal files, removing faces from photos for insurance claims, or removing account numbers from bank statements for loan brokers.

## What could go wrong for Redact PDF AI?

- **Revenue data stopped on 11 October 2026.** TrustMRR's Stripe API key expired on 11 October 2026. The figures on this page are the last synced values; the zeros shown for later days mean no data, not no sales. A reader should check whether the profile has been reconnected.
- **Small and uneven revenue.** US$1,919 in total, with almost 29% of it in May 2026. Monthly revenue from July to September 2026 was about 31% below April to June. The MRR of US$350 at the last sync was higher than any month since May, so the latest subscriptions may change the picture.
- **Compliance words on the page.** The page title says "EU-Hosted, GDPR & HIPAA", the badges say "HIPAA-ready Azure" and "SOC 2 Azure", and the healthcare section speaks of "HIPAA-compliant redaction". The FAQ states that "Redact PDF AI itself is not independently audited for these frameworks" and that the certificates belong to Microsoft Azure. A buyer in healthcare should ask for the BAA that the security page says can be discussed.
- **Unchecked claims.** The home page shows "99.9%" accuracy and more than 10,000 pages redacted, and the founder's TrustMRR text repeats the 99.9% figure. No test or source supports the accuracy number. The page count is possible at the revenue shown, but it is not verified.
- **Cookies and advertising.** The security page says "No third-party advertising or tracking cookies are used", while the privacy text lists "Google Ads/gtag (site conversion measurement)" among the services, and the home page loads a Google Ads tag. The footer also links to Caviard AI, a tool to redact ChatGPT conversations; no source on this page says who owns it.
- **Free tools already exist.** Adobe Acrobat and macOS Preview can redact PDFs, and the company's blog explains how to use Preview. The product must stay faster and more accurate than these to keep paying users.

## Sources

1. TrustMRR profile: Redact PDF AI (Stripe-verified revenue, data synced to 11 October 2026) — https://trustmrr.com/startup/redact-pdf-ai.md (grade A). Read the markdown profile, live on 3 November 2026, and compared it with a copy downloaded on 31 October 2026; the revenue figures are the same. Grade A for the Verification Sources and revenue sections; the description and founding date are founder-written, grade C.
2. Redact PDF AI home page, with pricing and FAQ — https://www.redact-pdf.ai/ (grade C). Read in full on 3 November 2026, including the FAQ answers in the page source. Source for the product, prices, usage claims and the compliance statement.
3. Redact PDF AI security page, with the privacy text in its page data — https://www.redact-pdf.ai/security (grade C). Read in full on 3 November 2026. Source for hosting, AI services, retention, the subprocessors and the data controller's name.
4. Redact PDF AI blog index — https://www.redact-pdf.ai/blog (grade C). Read the list of posts and their dates. Source for the oldest post date, 6 December 2025, and the October 2026 posts.
5. Damien B., Redact PDF AI blog, 22 August 2026: How AI Agents Can Redact PDFs: API Guide + MCP (2026) — https://www.redact-pdf.ai/blog/how-ai-agents-redact-pdfs (grade C). Read in full. Founder's post. Source for the API flow, the keyless demo and the MCP server.
6. Database of failed startups, local export — /method/#dead (grade C). Local export. AI-written summaries, not independently verified. Source for the failed predecessors.