mcp-parsekit

MCP server for ParseKit — document intelligence API for AI agents.

Submit any PDF, scanned image, spreadsheet, or email to ParseKit and get back clean, structured JSON with per-field confidence scores. 50+ built-in document types: invoices, contracts, W-2s, receipts, passports, and more.

Installation

npm install -g mcp-parsekit

Or run directly:

npx mcp-parsekit

Configuration

Set your ParseKit API key as an environment variable:

export PARSEKIT_API_KEY=your-api-key-here

Get your key at parsekit.dev/dashboard.

Claude tool definition

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "parsekit": {
      "command": "npx",
      "args": ["mcp-parsekit"],
      "env": {
        "PARSEKIT_API_KEY": "your-api-key-here"
      }
    }
  }
}

Available tools

| Tool | Description | |------|-------------| | createParse | Submit a document (PDF, image, spreadsheet) for structured extraction | | listParses | List recent parse jobs, with optional status/type filtering | | getParse | Retrieve a parse result by ID — use to poll async jobs | | listDocumentTypes | Browse all 50+ built-in document types and their schemas | | getDocumentType | Get the full field schema for a specific document type | | createSchema | Save a reusable custom extraction schema | | listSchemas | List your saved custom schemas | | getSchema | Retrieve a saved schema by ID | | deleteSchema | Delete a saved schema | | createBatch | Submit up to 100 documents for parallel async processing | | getBatch | Check batch progress (total / completed / failed counts) | | listBatchParses | Retrieve all parse results within a completed batch | | createWebhook | Register a webhook to receive real-time parse completion events | | listWebhooks | List registered webhooks | | deleteWebhook | Remove a webhook endpoint |

Example usage

Parse an invoice from a URL:

Use ParseKit to extract the vendor, amount, and line items from this invoice:
https://example.com/invoices/q1-2026.pdf

Claude will call createParse with type: "invoice" and return structured JSON.

Process 50 receipts at once:

I have 50 receipt URLs. Batch process them all with ParseKit and summarize the total spend.

Claude will call createBatch, poll getBatch until complete, then call listBatchParses to aggregate the results.

Custom extraction schema:

Extract the vendor name, SLA uptime guarantee, and termination notice period from this contract PDF.

Claude will call createParse with a custom schema matching those exact fields.

Supported document types

invoice · receipt · contract · nda · w2 · bank_statement · pay_stub · purchase_order · drivers_license · passport · insurance_card · medical_claim · shipping_manifest · customs_declaration · resume · business_card · lease_agreement · meeting_minutes · email · spreadsheet · and 30+ more

Call listDocumentTypes to see the full registry.

Links

• ParseKit documentation
• OpenAPI spec
• Dashboard
• Support