Norman MCP Server provides a financial operations cockpit that can be accessed through any MCP-compatible AI, such as Claude, ChatGPT, Cursor, or n8n. It connects your accounting, invoicing, VAT filing, and other financial workflows directly to your AI agents so you can perform tasks like creating invoices, categorizing transactions, pulling financial overviews, and generating tax reports from within your conversational interface. The server exposes an MCP-compatible endpoint (as shown in the project docs) that your AI agent can query to read and mutate financial data via ready-to-use skills. These skills include a full financial overview, creating invoices, managing clients, tax reporting, transaction categorization, receipt matching, overdue reminders, and more. You can invoke these capabilities through your preferred MCP client by pointing them to the provided endpoint and using the defined skill names.

In practice, you’ll enable OAuth 2.1 authentication and connect Norman to your Norman account. Your AI agent can then request operations such as generating Finanzamt previews, syncing Stripe payments, uploading receipts, and sending payment reminders. The skills map directly to common accounting workflows: financial-overview for dashboards, create-invoice to step through invoice creation and dispatch, manage-clients for CRM-like tasks, tax-report for VAT filing, and monthly-reconciliation for end-of-month close. The server’s design emphasizes seamless integration with Claude Code, OpenClaw, and Agents Skills ecosystems, making it straightforward to augment existing workflows with financial intelligence.

To use it with your AI agent, configure the MCP client to target the Norman MCP endpoint (for example, https://mcp.norman.finance/mcp) and ensure your agent authenticates via OAuth. Then call the available skills by their names (e.g., financial-overview, create-invoice, categorize-transactions) to perform the desired actions. The result is a cohesive financial assistant experience embedded inside your AI workflow.

Prerequisites

• Node.js (recommended LTS) and npm/yarn installed
• Internet access to fetch the Norm an MCP server package
• An OAuth 2.1 client/credentials if using OAuth-based authentication

Step-by-step installation

1. Install Node.js and npm (if not already installed)

   • macOS (using Homebrew): brew install node
   • Windows: download from https://nodejs.org/
   • Linux (Debian/Ubuntu): sudo apt-get update sudo apt-get install -y nodejs npm

2. Run the Norman MCP Server using npx (no local install required)

   npx -y norman-finance/norman-mcp-server

3. (Optional) Install locally for faster start and offline use

   mkdir norman-mcp-server && cd norman-mcp-server npm init -y npm install norman-finance/norman-mcp-server --save

   Start locally (adjust path if a local script is provided by the package)

   npx norman-finance/norman-mcp-server

4. Configure OAuth and environment variables (placeholders)

   • Set your OAuth client details in your environment or a .env file, for example: OAUTH_CLIENT_ID=your-client-id OAUTH_CLIENT_SECRET=your-client-secret OAUTH_SCOPES=read write API_BASE_URL=https://mcp.norman.finance/mcp

5. Verify the server is running

   • Access the MCP endpoint in your browser or via your MCP client to ensure the service responds correctly.

Notes

• If you’re deploying to a hosting environment, consider using Docker or a process manager (PM2, systemd) to keep the MCP server alive.
• Replace placeholder values with your real OAuth credentials and API URLs before production use.

Tips and common issues:

• Ensure your OAuth 2.1 credentials are configured and valid; failed authentications will prevent access to financial data.
• If you don’t see expected skills in your MCP client, verify that the endpoint URL (https://mcp.norman.finance/mcp) is correctly configured and that the client has permission to access the Norman MCP server.
• For local testing, running via npx keeps dependencies lightweight, but a local install can provide faster startup and easier version control.
• If you encounter network or CORS issues, confirm the MCP endpoint is reachable from your environment and that any required headers are present.
• The server surface includes a set of ready-to-use skills such as financial-overview, create-invoice, manage-clients, tax-report, categorize-transactions, and find-receipts. Use these names when instructing your AI agents to perform tasks.
• Keep client secrets and tokens secure; avoid committing them to public repositories.
• When integrating with multiple AI agents, ensure you route requests to the correct MCP endpoint and that each agent uses appropriate OAuth credentials.