Extend MCP server
{
"mcpServers": {
"extend": {
"type": "http",
"url": "https://mcp.extend.ai/mcp"
}
}
}{
"mcpServers": {
"extend": {
"type": "http",
"url": "https://mcp.extend.ai/mcp"
}
}
}{
"mcpServers": {
"extend": {
"type": "http",
"url": "https://mcp.extend.ai/mcp"
}
}
}{
"servers": {
"extend": {
"type": "http",
"url": "https://mcp.extend.ai/mcp"
}
}
}{
"mcpServers": {
"extend": {
"type": "http",
"url": "https://mcp.extend.ai/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.0.0
What it does
Extend exposes a hosted MCP server with 80+ tools for processing unstructured documents (PDFs, images, Office files, spreadsheets) into structured data. It covers extraction, parsing, classification, splitting, PDF form filling, and multi-step workflows, and is aimed at developers building document-processing agents or integrations.
- Extract structured fields from PDF invoices or contracts by connecting an MCP client to the hosted Extend server via OAuth.
- Classify and split multi-document PDFs into labelled segments as part of an automated document workflow.
- Fill PDF forms programmatically by passing structured data through the server's PDF editing tools.
Tools this server exposes
| Tool | Does |
|---|---|
| extraction | Pull specified fields from unstructured documents and return them as structured data. |
| parsing | Parse document content (PDFs, images, Office files) into machine-readable representations. |
| classification | Assign category labels to documents or document segments. |
| splitting | Divide a multi-page or multi-document file into separate, labelled parts. |
| pdf_form_filling | Populate fillable PDF form fields with provided structured data. |
| workflows | Run multi-step document processing pipelines defined in the Extend platform. |
| files | Upload and manage document files used as inputs for processing tools. |
| webhooks | Register and manage webhooks for async document processing notifications. |
| evaluations | Assess extraction or classification results against expected outputs. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.