Claude Desktop

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

Hosted (recommended)

{
  "mcpServers": {
    "talonic": {
      "url": "https://mcp.talonic.com/mcp",
      "headers": {
        "Authorization": "Bearer tlnc_your_key_here"
      }
    }
  }
}

Local (npx)

{
  "mcpServers": {
    "talonic": {
      "command": "npx",
      "args": ["-y", "@talonic/mcp@latest"],
      "env": {
        "TALONIC_API_KEY": "tlnc_your_key_here"
      }
    }
  }
}

Fully restart Claude Desktop (Cmd+Q on macOS, not just close the window). Talonic appears in the connected servers list with all eight tools.

Claude Desktop is the most popular MCP client for Talonic. It natively supports drag-and-drop file uploads, which pairs well with talonic_extract via the file_data parameter. When you drop a PDF into the chat, Claude automatically base64-encodes it and passes it to the MCP tool.

After connecting, you can verify the setup by asking Claude to list your schemas (talonic_list_schemas) or extract data from a sample document. If the tools do not appear, check that your JSON is valid and the config file is in the correct directory for your operating system.

Verify the connection

After restarting Claude Desktop, open a new conversation and ask Claude to list your schemas. A successful response confirms the MCP server is connected and your API key is valid. You should see the Talonic tools in the connected servers panel on the left side of the Claude Desktop window.

// Ask Claude: "List my Talonic schemas"
// Claude calls talonic_list_schemas with: {}

// Expected response (new workspace):
{
  "schemas": []
}

// If the key is invalid, you will see:
{
  "error": "authentication_error",
  "message": "Invalid or revoked API key. Check TALONIC_API_KEY in your config."
}

Claude Desktop is the recommended client for document-heavy workflows because of its native file handling. When you drop a file into the chat, Claude reads the file bytes, base64-encodes them, and passes them to talonic_extract or talonic_to_markdown via the file_data parameter. This means you can go from a physical document scan to structured JSON in under ten seconds — drop the file, Claude picks the right tool, and the extracted data appears in the conversation.

For multi-document workflows, Claude Desktop maintains conversation context across tool calls. You can extract data from several invoices in sequence, and Claude remembers all previous results. Ask follow-up questions like 'which of those invoices had the highest total?' and Claude reasons over the extracted data without re-calling the tools. This makes Claude Desktop particularly effective for batch review and comparison tasks.