How to Use Mistral OCR 4.1 in Amp
Extract text, tables, and images from PDFs and images, with Markdown content and paragraph bounding boxes. Use it to turn public reports, research papers, or scanned documents into searchable text with locations you can check against the page.
Prompts to try with Mistral OCR 4.1 in Amp
Copy an example into Amp, or ask in your own words.
Use Mistral document text extraction on https://www.berkshirehathaway.com/letters/2024ltr.pdf and add the extracted Markdown to this project's document search fixtures.
Use Mistral document text extraction on https://arxiv.org/pdf/1706.03762 and build a table review screen that shows extracted cells beside their source page.
Use Mistral document text extraction on Berkshire Hathaway's 2024 shareholder letter at https://www.berkshirehathaway.com/letters/2024ltr.pdf and add clickable paragraph highlights to our PDF viewer.
Use Mistral document text extraction on https://arxiv.org/pdf/1706.03762 and write a document importer that preserves headings, equations, and page references from the returned content.
Use Mistral document text extraction on https://arxiv.org/pdf/1706.03762 and add a figure browser to this project using the extracted images and their page locations.
How to set up Mistral OCR 4.1 in Amp
Pick the way that suits you. Each takes about two minutes.
Use Spiral's MCP server
- Add the server
No header and no client ID or secret: Amp registers itself with Spiral.
amp mcp add spiral https://api.spiral.ai/mcp
- Sign in
Start amp. Amp opens Spiral's sign-in page in your browser: sign in, pick the organization and click Authorize. No API key.
- Check it
amp mcp doctor
Use the Spiral skill
- Send Amp this message
Your agent reads Spiral's skill and sets itself up.
set up https://spiral.ai/SKILL.md
- Sign in or paste a key
Approve Spiral in the browser, or give the agent a key from the console.
Use the Spiral CLI
- Install and sign in
Installs the Spiral CLI, signs you in and adds its skill to your coding agents.
npx -y spiral-cli@latest init
Call the Spiral API
- Create an API key
In the console, under API keys.
- Run Mistral OCR 4.1
One request; the result comes back by polling or a webhook.
curl https://api.spiral.ai/v1/runs \ -H "Authorization: Bearer $SPIRAL_API_KEY" \ -H "Content-Type: application/json" \ -H "Idempotency-Key: $(uuidgen)" \ -d '{ "provider": "mistral", "model": "mistral-ocr-4-1", "input": { "document": { "document_url": "<document_url>" } } }'
Then paste one of the prompts above: Amp finds Mistral OCR 4.1 on Spiral, shows the price and runs it.
Mistral OCR 4.1 providers and prices
Spiral runs Mistral OCR 4.1 through Mistral. Open a card for today's price, with every fee included.
Settings you can ask for
Mistral OCR 4.1 works from your document annotation prompt. Ask Amp for any of these in your prompt.
| Setting | Allowed values | Default | Description |
|---|---|---|---|
| Image limit | a whole number | Max images to extract | |
| Table format | markdown or html | ||
| Extract footer | on or off | off | Extract the page footer into the response's |
| Extract header | on or off | off | Extract the page header into the response's |
| Image min size | a whole number | Minimum height and width of image to extract | |
| Include blocks | on or off | on | Return paragraph-level bounding boxes for all content blocks in the response |
| Include image base64 | on or off | Include image URLs in response | |
| Confidence scores granularity | word, page, or block | Granularity for confidence scores: 'page' (aggregate only), 'word' (per-word scores), or 'block' (per-block scores). Defaults to None (no confidence scores) to keep response payload small. |
Every API field of Mistral OCR 4.1 is on its tool page: open its card above.
What can you do with Mistral OCR 4.1 in Amp?
Build a searchable report viewer with extracted text linked to its position on each PDF page. Mistral's paragraph bounding boxes and structural labels give the project source regions to display beside search results.
Search public annual reports
Convert annual report PDFs into searchable text, keeping page locations beside passages about operating results and business risks.
Extract research tables
Pull tables from research papers into a review worksheet, then compare the extracted cells with the original pages before analysis.
Build a document viewer with source highlights
Use paragraph bounding boxes to highlight the source region when a reader selects extracted text.
Compare editions of public reports
Extract each edition of a company report and compare passages or tables to identify changes for an analyst to review.
Prepare public documents for retrieval
Turn PDF collections into Markdown passages with document and page references for a search or question-answering system.
Create a figure review queue
Collect extracted figures from public research papers and keep them beside their source pages for captioning or editorial review.
Structure public procurement documents
Extract document titles, submission deadlines, and requirements from public tender PDFs into a consistent review record.
More on Mistral OCR 4.1 and Amp
Overview
Mistral OCR 4.1 in other agents
Questions about Mistral OCR 4.1 in Amp
Can Amp use Mistral OCR 4.1?
Yes. Add Spiral to Amp as an MCP server, and Amp can find and run Mistral OCR 4.1 like any built-in tool.
How much does Mistral OCR 4.1 cost in Amp?
$0.0052 per page ($0.0065 with document or bounding-box annotations). You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Amp plan is separate.
Do I need a Mistral account to use Mistral OCR 4.1 in Amp?
No. Spiral runs Mistral OCR 4.1 on its own Mistral account. You sign in to Spiral once and never paste a key.
How long does it take to set up Mistral OCR 4.1 in Amp?
About two minutes: add the server, then sign in, then check it.
Does Amp wait while Mistral OCR 4.1 runs?
Mistral OCR 4.1 runs in the background. Amp starts the run, checks on it and gives you the json when it's ready.
Does Mistral OCR retrieve the latest version of a report?
It extracts content from the PDF or image supplied for processing. The document determines the reporting period and freshness; processing an older report does not update its contents.
How large a document can Mistral OCR process?
Mistral's chunking cookbook lists typical limits of approximately 50 MB and 1,000 pages. For larger documents, split the PDF into smaller sections before extraction.
What language coverage does Mistral document?
Mistral documents OCR 4 support for 170 languages across 10 language groups. That coverage statement does not promise identical recognition accuracy across languages.
Can extracted text be located on the original page?
OCR 4.1 supports paragraph bounding boxes and structural block labels. Those locations let a document viewer highlight the source region beside the extracted text.
Does document annotation examine every extracted figure?
Mistral's documented annotation workflow sends the extracted Markdown and the first eight extracted image bounding boxes to a vision model. A document annotation therefore does not include every image in a report containing more than eight extracted images.
Start using Mistral OCR 4.1 in Amp
One Spiral account for every tool your agent needs. No subscription; you pay per run.
Paste this into Amp:
set up https://spiral.ai/SKILL.md


