frenchie

Frenchie — your agent's best friend.
Make your agent do more.

Frenchie is an MCP-first utility that makes your AI agent more capable. OCR for PDFs and images. Transcription for audio and video. File extraction, generation, and more — coming soon.

One MCP interface. Sync and async routing. Credit-based pricing. Files are processed and deleted — not stored. Your first 100 credits are free.

MCP tool call

> tool: ocr_to_markdown
  file: invoice.pdf (4 pages, 1.2MB)

  credits   4 (1/page)
  route     sync
  status    done — 2.1s
  result    markdown returned to agent

> tool: transcribe_to_markdown
  file: customer-call.wav (47:12, 38MB)

  credits   96 (2/min)
  route     async — 5 chunks
  job_id    fr_job_29xk
  status    processing...

> tool: get_job_result
  job_id: fr_job_29xk

  status    done — 4m 12s
  result    markdown returned to agent
  expires   30 min after first delivery
serviceonline
PDF / image OCR
audio / video transcription
MCP interface
pay-per-use credits

capabilities

Your agent's toolkit — and it's growing.

You don't want to build upload flows, chunking logic, provider retries, and cleanup jobs for every new capability your agent needs. Let Frenchie fetch.

01
OCR

Send a PDF or image. Get Markdown back. Frenchie handles page splitting, provider routing, and merge logic. You get one clean result — not 47 API calls and a headache.

02
Transcription

Send audio or video. Get a Markdown transcript back. Long files are chunked and reassembled automatically. Your agent gets text it can actually work with, not a binary blob.

03
MCP-first

Call Frenchie through MCP from any agent. Results come back as Markdown your workflow can use immediately. No SDK, no REST client, no "let me just write a wrapper real quick."

benchmark

Your tokens. Saved.

Native PDF reading burns tokens on image encoding. Audio? Your agent can't even open it. Frenchie handles both outside the context window.

OCR

Scanned contract (12 pages)

48.0k tokens
6.2k tokens

Product spec PDF (28 pages)

112.0k tokens
14.8k tokens

Handwritten notes (3 pages)

can't read
1.9k tokens

Invoice image (1 page)

4.2k tokens
0.6k tokens
Native Frenchie

Transcription

Team standup (15 min)

not supported
30 credits

Client call recording (47 min)

not supported
94 credits

Podcast episode (1 hr 20 min)

not supported
160 credits
Native Frenchie

try it

See it yourself. Right now.

Drop a PDF, image, or audio file below. No account needed. We'll parse it and show you the Markdown.

flow

The whole path is visible.

No black boxes. Here's exactly what happens when you send a file.

01

Upload file via presigned URL

Your agent sends the file. We generate a secure upload link. Nothing touches disk longer than it needs to.

02

Inspect file, estimate credits, validate balance

We check file type, count pages or duration, and make sure your balance covers it before we start.

03

Route to the right tool

PDFs and images go through OCR. Audio and video get transcribed. Large files are chunked automatically.

04

Return Markdown — expire payload after delivery

Your agent gets clean Markdown back. The file and result are deleted after delivery. Nothing lingers.

quick start

Three steps. That's it.

From zero to parsing in under three minutes. We timed it.

01

Create an account & grab your API key

$ open https://getfrenchie.dev/register

Sign up, copy your API key from the dashboard. Takes 30 seconds.

02

Install the MCP server & skills

// Add to your MCP config:
{
  "mcpServers": {
    "frenchie": {
      "command": "npx",
      "args": ["-y", "@lab94/frenchie"],
      "env": { "FRENCHIE_API_KEY": "fr_your_key_here" }
    }
  }
}
Claude Code → ~/.claude/settings.jsonCursor → .cursor/mcp.jsonVS Code → .vscode/mcp.json
// Install slash commands (auto-detects your agent):
$ npx @lab94/frenchie install

Adds /ocr, /transcribe, and /frenchie-status commands to Claude Code, Cursor, Codex, or Antigravity.

03

Send file in. Get Markdown out.

> "Parse this PDF and summarize the key findings"
> "Transcribe this meeting recording and extract action items"

Your agent calls Frenchie, gets Markdown back, and keeps working. That's the whole thing.

pricing

$1 = 100 credits.

That's a coffee. Probably less than the coffee you're drinking right now.

1 creditper page — ocr
2 creditsper minute — transcription
  • 100 free credits on signup — no card required
  • Pay as you go — minimum $1 top-up
  • No subscriptions, no monthly fees, no "contact sales"
  • Credits don't expire

faq

You're going to ask these anyway.

What file types do you support?

PDF, PNG, JPG, WebP for OCR. MP3, M4A, WAV, MP4, MOV, WEBM for transcription. If it's a file and it has content, we probably handle it. And more types are on the way.

How good is the OCR?

Good enough to read scanned contracts with tables, handwritten notes, and that one PDF your client exported from a fax machine in 2003. Seriously — try the free credits and compare.

Can my agent read PDFs natively? Why do I need this?

It can — but scanned PDFs come back empty, and large ones eat your tokens alive. Frenchie parses outside your agent and returns clean text. Fewer tokens, better results.

What happens to my files?

Processed and deleted. We don't store your files after delivery. The result payload expires 30 minutes after first retrieval.

Is there a free tier?

100 credits on signup, no credit card required. That's 100 pages of OCR or 50 minutes of transcription. Enough to know if it works for you.

I can build this myself.

You absolutely can. If you're parsing more than 5,000 pages a month, you probably should. For everyone else, that's 4-6 hours of setup and ongoing maintenance for something that costs a dollar.

What agents does it work with?

Anything that speaks MCP — Claude Code, Codex CLI, Cursor, Windsurf, Antigravity, and whatever ships next Tuesday. One interface, all agents.

What's coming next?

File extraction, image and video generation, Markdown-to-file conversion, and free skill packs. Frenchie keeps learning new tricks. Follow the changelog.

Your agent could do more.
Frenchie can help.

100 free credits. No card required. Parse something in the next three minutes.