Connect DocuClipper to AI Agents (MCP), Docs

The docuclipper-mcp package is a stdio server that implements the Model Context Protocol (MCP). It lets AI tools (Claude Desktop, Cursor, Continue, Cline, and others) convert bank statements and invoices through DocuClipper without any manual uploading.

Prerequisites

A DocuClipper account with a Personal Access Token. Get yours at Settings → API & Webhooks (/account?section=api). See API Access: Personal Access Tokens.
Node.js 18 or later (for npx).

Setup

No install is required. All clients use npx docuclipper-mcp as the command.

Claude Desktop (macOS / Windows)

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "docuclipper": {
      "command": "npx",
      "args": ["-y", "docuclipper-mcp"],
      "env": {
        "DOCUCLIPPER_PAT": "dcp_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Restart Claude Desktop. DocuClipper tools appear under the tools menu (hammer icon).

Cursor

Add to .cursor/mcp.json in your project, or globally at ~/.cursor/mcp.json:

{
  "mcpServers": {
    "docuclipper": {
      "command": "npx",
      "args": ["-y", "docuclipper-mcp"],
      "env": {
        "DOCUCLIPPER_PAT": "dcp_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Other MCP clients (Continue, Cline, etc.)

Use the same command: "npx" / args: ["-y", "docuclipper-mcp"] pattern. Pass DOCUCLIPPER_PAT as an environment variable. Consult your client's MCP configuration docs for the exact file path and format.

Available tools

Tool	What it does
`convert_bank_statement`	One-shot: accepts a base64 PDF, uploads it, extracts, polls until done, and returns transactions (JSON or CSV). Best for files under ~10 MB.
`download_transactions`	Fetches transactions from any already-completed job by ID (up to 10,000 rows).
`upload_url`	Gets a presigned S3 URL for uploading large files directly.
`convert_document`	Enqueues an extraction job for a document you already uploaded.
`get_job_status`	Polls a job by ID and returns its status and transaction count.

For large files (over ~10 MB) use the async path: upload_url → PUT to S3 → convert_document → get_job_status → download_transactions.

Example prompt (Claude Desktop)

Once connected, you can ask Claude things like:

"Extract the transactions from this bank statement PDF and give me a CSV."

Attach the PDF to the message. Claude calls convert_bank_statement, waits for processing, and returns the structured data.

Troubleshooting

"DOCUCLIPPER_PAT environment variable is required": set env.DOCUCLIPPER_PAT in your client config and restart the MCP client.

HTTP 401: the PAT is invalid, expired, or was revoked. Generate a new one at Settings → API & Webhooks.

HTTP 402 / OutOfCredits: your account is out of pages. Top up at Settings → Plan & Billing.

Tool times out: convert_bank_statement defaults to 300 seconds. For long jobs, pass timeoutSeconds: 600 or use the async upload path instead.