# Frenchie

> Your agent's best friend. An MCP-first capability layer for AI agents: OCR for PDFs and images, transcription for audio and video, and image generation from text prompts — three capabilities through one install.

Frenchie is a closed-source managed service built by LAB94. The primary integration path is stdio MCP (`npx @lab94/frenchie install --api-key fr_…`); HTTP is available at `mcp.getfrenchie.dev` for hosted agents. Pricing is pay-as-you-go: $1 buys 100 credits, 1 credit per OCR page, 2 credits per transcription minute, 20 credits per generated image, 100 free credits on signup, no card required, credits don't expire.

## Core pages

- [Homepage](https://www.getfrenchie.dev/): Product overview — OCR and transcription for AI agents via MCP
- [Pricing](https://www.getfrenchie.dev/pricing): Pay-as-you-go pricing with usage examples and billing FAQ
- [Documentation](https://www.getfrenchie.dev/docs): Install, invoke, troubleshoot, migrate
- [Invocation by agent](https://www.getfrenchie.dev/docs/tools): Per-agent install and usage guides
- [Troubleshooting](https://www.getfrenchie.dev/docs/troubleshooting): Symptom-first fixes for common errors
- [Migrate](https://www.getfrenchie.dev/docs/migrate): Version-to-version upgrade guide
- [Changelog](https://www.getfrenchie.dev/changelog): Release history
- [About LAB94](https://www.getfrenchie.dev/about): The studio behind Frenchie
- [Terms & Privacy](https://www.getfrenchie.dev/terms): Legal

## Agent integrations

- [Frenchie in Claude Code](https://www.getfrenchie.dev/docs/tools/claude-code): Slash commands work out of the box. Results auto-save.
- [Frenchie in Cursor](https://www.getfrenchie.dev/docs/tools/cursor): Natural language works. One setting toggle on first install.
- [Frenchie in Codex (Desktop, CLI, IDE)](https://www.getfrenchie.dev/docs/tools/codex): @frenchie, /frenchie, or natural language — all three work.
- [Frenchie in Google Antigravity](https://www.getfrenchie.dev/docs/tools/antigravity): Invoke MCP servers by name, not by skill name.
- [Frenchie in VS Code (GitHub Copilot)](https://www.getfrenchie.dev/docs/tools/vscode): Slash commands work once MCP is registered.
- [Frenchie in Claude Desktop](https://www.getfrenchie.dev/docs/tools/claude-desktop): Natural language. Requires --global install.
- [Frenchie in Windsurf](https://www.getfrenchie.dev/docs/tools/windsurf): Natural language. User-level install.
- [Frenchie in Gemini CLI](https://www.getfrenchie.dev/docs/tools/gemini-cli): Natural language. Terminal-launched, no PATH gap.
- [Frenchie in Zed](https://www.getfrenchie.dev/docs/tools/zed): Assistant panel, user-level install.

## Comparisons

- [Frenchie vs Marker](https://www.getfrenchie.dev/compare/marker): Marker is a library you run. Frenchie is a server your agent calls.
- [Frenchie vs LlamaParse](https://www.getfrenchie.dev/compare/llamaparse): LlamaParse fits the LlamaIndex stack. Frenchie fits the MCP ecosystem.
- [Frenchie vs Whisper](https://www.getfrenchie.dev/compare/whisper): Whisper is a transcription model you run. Frenchie is a managed MCP server.
- [Frenchie vs AssemblyAI](https://www.getfrenchie.dev/compare/assemblyai): AssemblyAI goes deep on audio intelligence. Frenchie keeps it narrow.
- [Frenchie vs Deepgram](https://www.getfrenchie.dev/compare/deepgram): Deepgram is built for realtime. Frenchie is built for batch async.
- [Frenchie vs Replicate](https://www.getfrenchie.dev/compare/replicate): Replicate is a model-hosting platform. Frenchie is an MCP tool your agent calls.
- [Frenchie vs Fal.ai](https://www.getfrenchie.dev/compare/fal): Fal.ai optimizes for low-latency image generation. Frenchie optimizes for agent ergonomics.

## Use cases

- [Meeting transcription for AI agents](https://www.getfrenchie.dev/use-cases/meeting-transcription): Drop a recording. Get a clean transcript back. Let your agent summarize, extract decisions, or answer follow-ups directly.
- [Sales call transcription inside your CRM workflow](https://www.getfrenchie.dev/use-cases/sales-call-transcription): Frenchie transcribes sales calls into clean Markdown so your agent can extract next steps, objections, and commitments without relying on your memory or your rep's notes.
- [Podcast transcription for show notes, SEO, and accessibility](https://www.getfrenchie.dev/use-cases/podcast-transcription): Drop an episode. Get Markdown your agent can turn into chapter markers, pull-quote cards, SEO-ready show notes, and full accessibility transcripts.
- [Research paper parsing — text, tables, and figures](https://www.getfrenchie.dev/use-cases/research-papers): Frenchie parses PDF research papers into clean Markdown, pulling figures out as separate PNG files so your agent can cite specific content instead of summarizing the whole thing.
- [Contract and legal document OCR for agent review](https://www.getfrenchie.dev/use-cases/legal-documents): Frenchie parses legal PDFs — scanned or native — into clean Markdown so your agent can flag clauses, extract dates, and diff redlines without you scrubbing formatting by hand.
- [Invoice OCR and receipt extraction for bookkeeping agents](https://www.getfrenchie.dev/use-cases/invoices): Frenchie parses invoices and receipts into structured Markdown so your agent can extract line items, totals, tax, and dates straight into your accounting tool.
- [Handwritten notes to searchable Markdown](https://www.getfrenchie.dev/use-cases/handwritten-notes): Frenchie reads handwritten pages — notebook scans, whiteboard photos, meeting jottings — and returns Markdown your agent can search, summarize, and reference.
- [Product mockups for AI agents](https://www.getfrenchie.dev/use-cases/product-mockups): Describe a product image. Your agent generates it, saves it next to your work, and moves on.
- [Blog covers and social images for AI agents](https://www.getfrenchie.dev/use-cases/blog-covers): Your agent drafts a blog post, then generates a matching cover image in the same flow. No separate design pass.

## Blog

- [Read, listen, create: why your AI agent needs all three](https://www.getfrenchie.dev/blog/read-listen-create): Your agent already reasons. Give it three hands: read files, listen to recordings, create images. How image generation completes the Frenchie toolkit.
- [Why we built Frenchie: the MCP tools gap](https://www.getfrenchie.dev/blog/why-we-built-frenchie): Agents write code and reason about systems. Drop a scanned PDF or a 30-minute recording in front of one and most hit a wall. That wall is why Frenchie exists.
- [How Frenchie handles 30-minute audio without blocking your agent](https://www.getfrenchie.dev/blog/async-transcription-without-blocking): Sync transcription freezes your agent for five minutes. Async job handling is the detail that makes transcription usable inside a live agent workflow.
- [Inside Frenchie: from scanned PDF to clean Markdown in 3 seconds](https://www.getfrenchie.dev/blog/inside-frenchie-pdf-to-markdown): What happens between dropping a scanned contract into your agent and getting back searchable Markdown with tables and figures intact. The OCR pipeline.
- [MCP for beginners: what it is and why Frenchie picks it](https://www.getfrenchie.dev/blog/mcp-for-beginners): Model Context Protocol is the quiet standard reshaping how AI agents get tools. What it actually is, why it matters, why Frenchie bet the whole product on it.
- [Pay-per-use vs subscription: why Frenchie ships flat $1 = 100 credits](https://www.getfrenchie.dev/blog/pay-per-use-vs-subscription): Every SaaS eventually debates usage-based vs subscription pricing. For a narrow tool like Frenchie, the answer was clear before the first line of billing code.

## Key facts

- Built on: Model Context Protocol (MCP)
- Transports: stdio (local, primary) + HTTP (hosted, fallback)
- Tier-A auto-install: Claude Code, Cursor, Codex, Antigravity, Claude Desktop. Other MCP clients can use the same skill pack with manual config.
- Capabilities: OCR (PDF/image → Markdown), transcription (audio/video → Markdown), image generation (text prompt → PNG/JPEG/WebP via gpt-image-2)
- OCR formats: PDF, PNG, JPG, WebP
- Transcription formats: MP3, M4A, WAV, MP4, MOV, WebM
- Image generation output formats: PNG (default), JPEG, WebP. Sizes: 1024x1024, 1536x1024, 1024x1536, or auto. Quality: low/medium/high/auto. Background: transparent/opaque/auto (transparent rejected with JPEG).
- Image generation rate limits: 50 images/hour and 250 images/day per user.
- Max file size: 2 GB per file
- Result retention: 30 minutes after first delivery, then deleted
- Languages: 50+ including Thai, Japanese, Chinese, Arabic, French, Spanish
- Pricing: $1 = 100 credits; 1 credit/OCR page; 2 credits/transcription minute; 20 credits/generated image; 100 free credits on signup
- Shared result contract: every capability returns `{ status, jobId, creditsUsed, resultExpiresAt, result: { kind: "markdown" | "image", ... } }`. Agents branch on `result.kind`.
- Contact: support@getfrenchie.dev
- npm package: https://www.npmjs.com/package/@lab94/frenchie

## Optional

- [Full site content](https://www.getfrenchie.dev/llms-full.txt): Every page and blog post aggregated into a single Markdown document for full-context AI ingestion
- [RSS feed](https://www.getfrenchie.dev/blog/feed.xml): Blog updates in RSS 2.0
- [Sitemap](https://www.getfrenchie.dev/sitemap.xml): Canonical URL index