# PDF to Markdown AI

Aiolum story: https://aiolum.com/story/pdf-to-markdown-ai/ · added 2026-11-12 · updated 2026-11-12

**In short.** PDF to Markdown AI (PDF to Markdown converter with OCR, an API and an MCP server for AI agents, United States): US$758 all-time verified revenue, October 2026. Evidence grade A: disclosed by a company or a filing, or verified by a payment processor. Team: Built by an agency whose founder is named on the About page; headcount not published. Launched March 2025; still running, revenue falling.

PDF to Markdown converter with OCR, an API and an MCP server for AI agents · United States · AI-native · team of 2 to 10 · still running, revenue falling · group: Documents, data and slides

- Outcome: US$758 all-time verified revenue, October 2026 (evidence grade of the headline number: A; headline in US dollars: 758)
- Launch: March 2025; months from launch to the outcome: 19
- Team: Built by an agency whose founder is named on the About page; headcount not published
- Outside funding: None reported
- Main factor: Niche
- Meters, 1 to 4: evidence strength 3, copyable pattern 3, capital needed 1
- Tags: AI-native, Small team, Bootstrapped, Running, B2B, Information Technology

Links:

- Visit pdftomarkdown.ai: https://pdftomarkdown.ai/
- Verified revenue on TrustMRR: https://trustmrr.com/startup/pdf-to-markdown-ai

**The product.** A hosted web tool that turns a PDF into Markdown, a plain-text format with headings, lists and tables. It runs OCR (optical character recognition, which reads text from scanned pages) and extracts images. It also offers a REST API and an MCP server, so that AI agents such as Claude and Cursor can call it. Customers pay per page.

**Why it worked.** A narrow job with no subscription brought US$758 in total. The site charges per page from prepaid credits that never expire. It also sells to AI agents through an API and an MCP server. Monthly revenue peaked at US$330 in August 2026 and was US$66 in September.

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; E = no revenue figure found, so the headline is a non-money fact and the other facts carry their own grades; calc = computed on this page from graded facts.

## Summary

PDF to Markdown AI is a hosted converter. A user uploads a PDF and receives Markdown, a plain-text format that marks headings, lists and tables with simple characters. AI tools read it more reliably than a PDF. The site also offers a REST API and an MCP server, a small program that lets an AI agent call a tool. The site says the conversion runs on the API of a company called Datalab.

The About page says the product was built by Nathan Parikh, founder of HappyWP, a web and digital strategy agency in Fort Worth, Texas. TrustMRR gives 8 March 2025 as the start date. The Internet Archive's first saved copy of the domain has the same date. The verified payment records show US$758 over all time, US$620 in the last twelve months and US$92 in the last 30 days. Recurring revenue is US$0, because there is no subscription.

The headline figure is grade A. TrustMRR reads it from Stripe through an API key, and the profile was synced on 12 October 2026. The founder or seller wrote the start date, the country and the user estimate of 34,000 on the profile, so those are grade C. The sources do not give the size of the team.

## Why did PDF to Markdown AI work?

### It sells one narrow job to people who feed documents to AI

The home page, the llms.txt file and the About page all describe one job. That job is to turn a PDF into clean Markdown for tools such as Claude, ChatGPT, Cursor and Obsidian. The sitemap lists use-case pages for each of these tools, and pages on lecture notes, medical PDFs and legal PDFs. No source says how many customers each page brought.

### It charges per page with credits and no subscription

A user pays for pages, not for a month. Bundles run from US$10 for 400 credits to US$375 for 15,000 credits, and credits never expire, the site says. TrustMRR shows US$0 of recurring revenue and no active subscription.

### It offers the converter to AI agents through an API and MCP

The llms.txt file gives a REST endpoint and an npm package, @happywp/pdftomarkdown-mcp, for Claude Desktop and similar tools. The npm registry lists version 1.0.0 with an MIT licence. The sources do not say how much revenue the API or the package brought.

### Revenue came in bursts and fell after August 2026

Monthly revenue was US$3 in November 2025 and US$39 in January 2026. It rose to US$330 in August 2026 and fell to US$66 in September. The sources give no reason for the peak or the fall.

## How it was built

These details come from the site's page code and its own text. They are grade C unless marked otherwise.

- The conversion itself runs on Datalab's Marker API. The site says the file content is sent to Datalab for that purpose.
- Optional AI cleanup may send the Markdown to Anthropic or OpenAI, only when the user turns the feature on.
- The front end is a single-page web app written with React 18.3.1, as the public script shows. The home page loads Google Tag Manager and Google Analytics.
- Payments run through Stripe Checkout, the site says. TrustMRR reads the revenue from Stripe through an API key.
- The API takes a multipart file upload with a bearer key. The MCP server is an npm package for Claude Desktop, Cursor and similar tools.
- The site says original PDFs are not stored permanently. Signed-in users can keep, delete or switch off the history of converted Markdown.

Sources: [3, 8, 2, 1, 10]

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

Three companies in our database of failed startups converted files, collected web content, or built a tool on top of a platform that later absorbed it. 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.

### Markupwand (2012 to 2015; funding US$120K; cause: Market fit)

- Promise: A tool, backed by Y Combinator, that converted design files into clean HTML and CSS code.
- Why it died: The record says design-to-code was not a standalone solution for most companies. It adds that design tools such as Figma and Adobe now offer integrated solutions.
- Record: 

### DataNitro (2012 to 2015; funding US$120K; cause: Platform dependence)

- Promise: Run Python scripts inside Excel, so that analysts could automate their work without leaving it.
- Why it died: The record says the offer was limited by Excel's architecture, and Excel has since added Python itself.
- Record: 

### Springpad (2008 to 2014; funding US$7.5M; cause: Monetization)

- Promise: A tool to collect, organize and find web content such as articles, videos and recipes.
- Why it died: The record says the company could not turn its technology into a business, and its way of earning money did not use its user base.
- Record: 

## What changed

Two of the three sold a tool that a larger platform later offered itself: DataNitro next to Excel, and Markupwand next to the design tools. Springpad, its record says, could not turn its technology into a business. PDF to Markdown AI carries the first exposure. Its converter runs on another company's API, and AI model makers add document reading to their own products.

Two things differ. The output goes to AI tools that many people now use. The customer buys pages when needed and pays no fixed monthly fee. The cost of that choice is visible in the figures. Revenue stays small and has to be earned again each month.

## Can one person copy PDF to Markdown AI now?

**Verdict: The structure, yes. The size of the revenue, not yet.**

A hosted wrapper around a document-conversion API earned US$758 in total in 19 months. A copier should plan for a larger recurring part, since this site had US$0 of recurring revenue in October 2026.

1. Pick one document format that AI users struggle with, and name the AI tools in the page titles.
2. Sell pages or credits, as this site does, and also offer a plan that repeats each month.
3. Publish an API and an MCP server early, so that agents can call the tool without a person.
4. Write one page per tool and per use case, and check later which pages bring paying users.

**What is still open.** Other file formats that AI agents read badly: spreadsheets with merged cells, scanned forms, slide decks and engineering drawings.

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

A small verified total: US$758 in all, with a best month of US$330 in August 2026 and no recurring revenue

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Verified revenue, all time, to 12 October 2026 | US$758 | A | TrustMRR markdown profile, a copy cached in October 2026, revenue synced 12 October 2026. The data comes from Stripe through an API key. | [1] |
| Verified revenue, last 12 months | US$620 | A | Same profile. The monthly table runs from November 2025 to October 2026 and adds up to US$620. | [1] |
| Revenue before November 2025 | US$138 | calc | Calculated: US$758 minus US$620. The profile gives no month-by-month table before November 2025. | [1] |
| Best month in the table | US$330, August 2026 | A | The profile lists US$136 for July 2026, US$66 for September 2026 and US$36 for October 2026 up to the sync on 12 October. | [1] |
| Share of all revenue earned in August 2026 | about 44% | calc | Calculated: US$330 divided by US$758. | [1] |
| Revenue, last 30 days | US$92 | A | TrustMRR snapshot, with growth of +15.0% against the 30 days before. The daily table for the same period adds up to US$82. | [1] |
| Recurring revenue and subscriptions | US$0 a month, 0 active subscriptions | A | TrustMRR metric snapshot, October 2026. The site sells credits and metered billing, not a subscription. | [1, 3] |
| Start of the product | March 2025 | B | The Internet Archive's first saved copy of the domain is dated 8 March 2025. TrustMRR gives 8 March 2025 in its field "Founded date", which the founder wrote. No earlier name or domain was found. | [1, 7] |
| Maker | Nathan Parikh, founder of HappyWP | C | The About page says he built the product. It also says the product belongs to HappyWP, a web and digital strategy agency in Fort Worth, Texas. TrustMRR names the same founder. The sources do not give the year HappyWP started; its own site answered HTTP 403 on 12 November 2026. | [3, 1] |
| Team | headcount not published | C | The About page says "we" and gives no number of people. No source suggests more than 10. | [3] |
| How conversion works | Datalab's Marker API | C | The site says it is a hosted product layer powered by Datalab's document-intelligence API. It lists Anthropic and OpenAI for optional AI cleanup, only when the user turns it on. | [3] |
| Prices, November 2026 | US$10 for 400 credits up to US$375 for 15,000 credits; US$0.03 a page pay-as-you-go | C | Calculated from the price data: US$0.024 to US$0.026 per credit. | [4, 3] |
| Counter on the home page | 66,669 pages and 6,369 documents converted | C | The site's own statistics endpoint, read 12 November 2026. The home page shows the page count. | [5, 3] |
| Estimated user count | 34,000 | C | A field of the TrustMRR profile written by the founder or seller. It is larger than the 6,369 documents counted by the site. | [1, 5] |
| Pages on the site | 66 addresses in the sitemap | C | Pages for converters, comparisons with other tools, use cases and AI agents. The sitemap index is dated 20 July 2026. | [9] |

How to read this table. Grade A is a value that TrustMRR reads from Stripe through an API key. Grade B is a record kept by an independent party. Grade C is a value on the company's own pages or in the founder's profile fields. Grade X was computed on this page.

## Pricing

PDF to Markdown AI prices read on 12 November 2026. The bundles come from the site's price data and the other rows from its script (grade C)

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free | 0 | 0 | 10 pages without an account; the site also says 30 free pages in total |
| Top up, small | 10 one-time |  | 400 credits |
| Top up, medium | 36 one-time |  | 1,500 credits |
| Top up, large | 130 one-time |  | 5,000 credits |
| Top up, largest | 375 one-time |  | 15,000 credits |
| Pay-as-you-go |  | 0.03 per page | metered, billed monthly |

The pricing page loads its prices by script. The bundle prices come from the site's own price data, which is not shown as text on the page. One credit is one standard page. Optional AI cleanup costs 2 credits per page. Credits never expire, and unused credits are refundable within 30 days, the site says.

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

From a first saved page in March 2025 to a profile synced in October 2026. Each entry carries its grade.

- **8 March 2025** (grade B [7, 1]): The Internet Archive saves the first copy of pdftomarkdown.ai. TrustMRR gives the same day as the start date.
- **9 March 2025** (grade C [6]): An archived copy of the home page is titled "PDF to Markdown Converter" and says "Try for free".
- **November 2025** (grade A [1]): The first month in the monthly table: US$3 of verified revenue.
- **January 2026** (grade A [1]): US$39 of verified revenue. No month between February and May appears in the table.
- **August 2026** (grade A [1]): The best month, US$330, after US$136 in July.
- **September to 12 October 2026** (grade A [1]): US$66 in September and US$36 in October up to the sync. The profile syncs on 12 October with US$758 over all time.

## What could go wrong for PDF to Markdown AI?

- **A wrapper around another company's API.** The site says it is a hosted product layer powered by Datalab's API. Datalab also maintains open-source document projects such as Marker, which anyone can run locally. A customer who can run those tools has less reason to pay for this site.
- **Counters and badges on the home page without a source.** The home page shows "pages converted", "40+ languages supported" and "99.9%" uptime. The page count comes from the site's own statistics endpoint. No source confirms the language count or the uptime. No logo wall and no named customer testimonial was found in the site's script.
- **Figures that do not agree with each other.** The comparison tables say "From $0.003/page", but the price data gives US$0.024 to US$0.026 per credit. The pages say both 10 free pages and 30 free pages in total. The daily revenue table adds up to US$82, while the 30-day snapshot says US$92. The profile's 34,000 users stand next to 6,369 documents converted.
- **Agency owner, student pages and uploaded documents.** The maker, HappyWP, calls itself an agency, but the product is software that customers use themselves. The year the agency started was not found, and the product is counted from its own first saved page of March 2025. Some pages address students, with lecture slides, study notes and practice exams, and others address medical and legal PDFs. The terms say users may not process documents they have no right to access. Uploads go to Datalab for processing.
- **Founder-written profile fields.** The start date, the country, the founder's name and the 34,000 user estimate are fields on the TrustMRR profile. Only the revenue and subscription figures come from Stripe. The live TrustMRR profile answered HTTP 500 on 12 November 2026, so the cached copy of October 2026 was used.

## Sources

1. TrustMRR: PDF to Markdown AI, markdown profile (cached copy of October 2026) — https://trustmrr.com/startup/pdf-to-markdown-ai.md (grade A). Read in full from the pipeline's cached copy of October 2026. The live profile answered HTTP 500 on 12 November 2026. Revenue figures come from Stripe.
2. PDF to Markdown Converter for Claude & AI | PDFtoMarkdown.ai (home page) — https://pdftomarkdown.ai/ (grade C). Read in full on 12 November 2026. The text loads by script, so this copy gives the title, the description and the page code.
3. PDFtoMarkdown.ai site script (about, pricing, terms and FAQ text) — https://pdftomarkdown.ai/assets/index-CFMr_3rk.js (grade C). Searched the whole file on 12 November 2026. Source for the About page, the Datalab statement, the prices, the free pages, the terms and the home page counters.
4. PDFtoMarkdown.ai pricing tiers (the site's price data) — https://pdftomarkdown.ai/api/pricing-tiers (grade C). Read in full on 12 November 2026. Source for the four credit bundles and their prices.
5. PDFtoMarkdown.ai statistics (the site's counter data) — https://pdftomarkdown.ai/api/stats (grade C). Read in full on 12 November 2026. Source for the pages and documents converted.
6. pdftomarkdown.ai, Internet Archive copy of 9 March 2025 — https://web.archive.org/web/20250309101121/https://pdftomarkdown.ai/ (grade C). Read the title and the description. The page body loads by script and is not in the copy.
7. Internet Archive capture list for pdftomarkdown.ai — https://web.archive.org/cdx/search/cdx?url=pdftomarkdown.ai&limit=5&output=json&filter=statuscode:200 (grade B). Read in full on 12 November 2026. The first saved copy is dated 8 March 2025.
8. PDFtoMarkdown.ai llms.txt — https://pdftomarkdown.ai/llms.txt (grade C). Read in full on 12 November 2026. Source for the API endpoint, the MCP package and the use cases.
9. PDFtoMarkdown.ai sitemap.xml and sitemap index — https://pdftomarkdown.ai/sitemap.xml (grade C). Read in full on 12 November 2026. Source for the 66 addresses and the date of 20 July 2026 in the sitemap index.
10. npm registry: @happywp/pdftomarkdown-mcp — https://registry.npmjs.org/@happywp%2fpdftomarkdown-mcp (grade B). Read in full on 12 November 2026. Source for the version, the licence and the description of the package.
11. Database of failed startups, local export — /method/#dead (grade C). Record pages read on the website on 12 November 2026. AI-written summaries, not independently verified. Source for the failed predecessors.

Contact: aiolum@livyden.com. Run PDF to Markdown AI? Send corrections or a verified figure to aiolum@livyden.com (subject: Correction: PDF to Markdown AI).