For AI Agents

Give agents content tools they can chain

MCP agents need more than one-off tools. They need structured I/O, confidence scores, predictable errors, and compatible content outputs so a document extraction can feed a report, spreadsheet, image, or follow-up tool call.

Set Up the MCP Server

Zero data retention Made & hosted in the EU 100 trial credits

One output feeds the next

APIs that cover every step of EU content workflows — ingestion, transformation, generation, and exports. One key, one credit pool, and structured JSON responses designed to chain together.

Extract inputs

Turn documents, websites, and other sources into structured output.

Document Extraction

Extract structured data from documents with confidence scores.

Document to Markdown

Convert documents, images, and websites into clean Markdown.

Website Extraction

New

Extract typed JSON from public web pages with citations.

Audio Extraction

Coming Soon

Transcribe audio and extract structured data with speaker labels.

Generate outputs

Create images, spreadsheets, and files from the same workflow state.

Image Transformation

Resize, crop, upscale, and convert images in one request.

Image Generation

Compose dynamic images from layers, text, QR codes, and fonts.

Document Generation

Generate PDF, DOCX, EPUB, and PPTX files from JSON content.

Sheet Generation

Generate XLSX and CSV sheets with formulas and styles.

Fits into your existing stack

Native SDKs for TypeScript, Python, and Go. OpenAPI spec for everything else. MCP server for AI agents and Claude Code skills. n8n integration for visual workflows.

EU AI workflow runtime

Run document, image, and file steps through one EU-hosted workflow layer with shared API conventions and billing.

You're dealing with this every day

Here's what changes when you stop juggling tools and start building pipelines.

Outputs agents cannot trust

Most content APIs return blobs, vague failures, or fields with no confidence scores. Agents need structured I/O and predictable errors before they can run without constant human repair.

Structured, typed I/O

JSON schemas in, typed fields out. Document Extraction includes confidence scores, and every tool returns structured data agents can inspect before choosing the next tool.

One vendor per capability

Your agent needs to read documents, extract websites, process images, and generate files, but each capability often comes from a different service with different auth, schemas, and failure modes.

Every tool, one server

Your agent sees document extraction, document-to-markdown, website extraction, image transformation, document generation, image generation, and sheet generation through one MCP server.

MCP setup without a workflow

Many MCP servers expose a narrow action. Content workflows need a coherent tool set: extract, transform, generate, and pass the output into the next step through one server.

Interactive now, production later

Prototype in Claude, Cursor, or Windsurf over hosted MCP. When the workflow graduates, the same operations are available through REST and the TypeScript, Python, and Go SDKs.

Real-world pipelines, ready to ship

Each recipe chains multiple APIs into a complete workflow. Pick one, tweak it, and deploy — or use it as a starting point for your own pipeline.

Compress Image to Target File Size

Compress an image to fit within a specific file size in bytes using quality-first compression.

Convert Document for Knowledge Base

Convert external documents — specs, contracts, reports — to markdown for knowledge base ingestion.

Convert Document for RAG Ingestion

Convert a document to clean markdown suitable for chunking and embedding in a RAG pipeline.

Convert Image Format

Convert an image between PNG, JPEG, and WebP formats with quality control for web optimization.

Convert Website to Markdown for RAG

Extract clean title, summary, markdown sections, and source metadata from a public documentation page for RAG ingestion.

Convert Document to Markdown

Convert PDF, DOCX, HTML, or image documents to clean, structured Markdown.

Browse all recipes

Official SDKs for every major language

Install the SDK, set your API key, and start chaining requests. Full type safety, automatic retries, and idiomatic error handling included.

TypeScript Python Go

European by design

Your data is processed on EU-hosted infrastructure and never stored beyond temporary logs. Zero data retention, GDPR-compliant workflows, and a Data Processing Agreement are available for every customer. Learn more about our security practices .

EU-hosted core processing

Application and processing infrastructure runs in Europe, with provider-scope ISO 27001 and BSI C5 evidence documented for procurement reviews.

Zero data retention

Customer files and processing results are not stored after the request. Usage logs are retained for 90 days and automatically deleted.

Clear answers for security teams

Give reviewers the answers they need up front: where files are processed, what is retained, which subprocessors are involved, and how AI inputs, outputs, review gates, and audit records move through each workflow.

Trust and Compliance EU Hosting Responsible AI Use Security Controls

Pricing

Start usage-based. Switch to a subscription when your volume becomes predictable.

Pay as you go

Usage-based

$0.099 to $0.069 / credit

Graduated pricing. Your effective rate decreases automatically as monthly usage grows.

No monthly commitment
Pay only for credits used
Automatic volume discounts as usage grows

Start 7-day trial

Subscriptions

Predictable volume

From $19 /month

Fixed recurring credit packs with lower effective credit prices for steady usage.

Lower effective per-credit prices
Fixed recurring credit packs
Predictable monthly budget

Compare Plans

Trial Week

Try Iteration Layer with 100 credits

$1 / 7 days

Start with one shared trial pool before choosing subscription or pay-as-you-go billing.

Start 7-day trial

$1 / 7 days

All APIs included 7-day trial Project-based budget caps Auto overage billing

Compare pricing options

Solving real problems, step by step

Vendor sprawl, GDPR headaches, broken pipelines — we wrote about the problems you're facing and how to fix them.

The Model Is Not the Moat. The Orchestration Layer is.

Model choice is becoming interchangeable for many agent workflows. Durable systems depend on schemas, routing, review, state, and outputs.

Building AI Agents That Process Documents: MCP, Structured I/O, and Confidence Routing

Build an AI agent pipeline that extracts document data, evaluates confidence scores, and routes to report generation or human review — using MCP and composable APIs.

MCP First, REST Later: How AI Workflows Mature into Production Pipelines

Use MCP to discover the workflow, then move stable document and image processing paths into REST, SDKs, or automation.

Frequently asked questions

How do I connect the MCP server to my agent?

Add the hosted Streamable HTTP endpoint to your MCP client's settings. The server uses OAuth 2.1, so your client opens a browser authorization flow instead of asking for an API key. See the MCP setup docs.

Can agents chain multiple operations?

Yes. An agent can extract data from a PDF, check confidence scores, generate a review document for low-confidence fields, and export approved rows to a spreadsheet through the same MCP server.

What is a concrete agent workflow?

Invoice intake: the agent extracts vendor names, invoice numbers, line items, totals, and due dates; routes low-confidence fields for review; generates a PDF summary for approved invoices; and writes the batch to an XLSX file for accounting.

What content types can agents process?

Documents (PDF, DOCX, XLSX, CSV), images (PNG, JPEG, WebP, HEIF), HTML, Markdown, text, and public web pages. Agents can extract data, transform images, and generate PDFs, spreadsheets, and images through the same MCP server.

What response format do the APIs return?

Every API returns structured JSON with typed fields. Document Extraction returns field-level data with confidence scores. Image Generation returns base64 or URLs. All responses follow consistent schemas your agent can parse without guesswork.

Do you have Claude Code skills?

Yes. We publish Claude Code skills that wrap our APIs into ready-to-use agent capabilities. See the AI Integration docs.

Do you provide an OpenAPI spec?

Yes. A full OpenAPI 3.1 specification is available at our gateway endpoint. Use it for code generation, IDE integration, or to build custom agent tool definitions.

How do agents handle errors?

All tools return structured JSON errors with a consistent format: error type, message, and HTTP status. No HTML error pages or ambiguous responses. Credits are refunded on server errors automatically. See Errors.

Is there rate limiting?

Yes, but generous. Rate limits are per API key and designed for production workloads. If your agent needs higher limits, contact us — we can adjust per account. See Rate Limits.

Build your first workflow in minutes

Chain our APIs into a workflow you can test with your own data during the 7-day trial.

Start 7-day trial

Zero data retention Made & hosted in the EU 100 trial credits