Integrations for AI assistants and automations

Before you start

All integrations use an API key from your account. Create a test key first (free, a few calls per day) and switch to a normal key when everything works. Calls are charged to your account like on the website: the daily free quota first, then credits; failed calls cost nothing.

Create an API key API documentation

Claude and ChatGPT: connect your account

In Claude and ChatGPT you add Wandlix as your own connector without copying an API key: you sign in once with your e-mail address and the code and allow access. Afterwards you simply say in the chat, for example, “compress this PDF” or “check this e-invoice”.

Server address: https://mcp.wandlix.com/mcp (equivalent: https://wandlix.com/api/v1/mcp)

1. Add the connector

   In Claude (claude.ai or desktop) open the settings under “Connectors” and choose “Add custom connector”; in ChatGPT use the connectors or apps settings to add your own MCP server (depending on your plan in developer mode). Enter the server address above.

2. Sign in and allow

   A page of Wandlix opens. Sign in with your e-mail address and the 6-digit code and click “Allow”. There is no password.

3. Use it in the chat

   Attach the file if your assistant can, or describe the task. Your daily free allowance is used first, then your credits. Ask the assistant for your balance; the tool for it is called get_balance.

How do files go in and out?

Small files go into the call as Base64. For larger files, or files in the assistant’s code environment, the assistant orders a one-time upload link (tool create_upload) and uploads the file there. Every result is also available under a one-time download link. All links are valid for 60 minutes at most, files are kept in memory only and deleted after they are fetched. In Claude the code environment must be allowed to reach the domain wandlix.com: Settings, Capabilities, allowed domains, add wandlix.com. If it is blocked, the assistant sends small files (up to about 1 MB) as Base64 directly in the call.

Important for privacy

If you use Wandlix through an AI assistant, your files and requests also pass through that assistant’s provider (for example Anthropic or OpenAI). Their terms and privacy notices apply to that part. Wandlix itself does not store files; only the name of the application, timestamps and the checksum of the access tokens are stored.

Disconnect connections in your account · Privacy policy · support@wandlix.com

MCP for AI assistants

MCP (Model Context Protocol) lets an AI assistant call our tools as functions: “compress this PDF”, “check this e-invoice”, “convert this Word file”. Every Wandlix tool is available.

Server address: https://wandlix.com/api/v1/mcp

• Transport: HTTP (JSON-RPC); no extra software is needed on our side.
• Every request needs a sign-in: either through the connector (OAuth, see above) or your API key in the Authorization header. Without it the server answers 401 and points to the sign-in.
• Files go in as Base64 or through an upload link. Results up to 1 MB also come back as an embedded resource (several result files as a ZIP); every result has a one-time download link.
• Files stay in memory only and are deleted after they are fetched, after 60 minutes at the latest.

Claude Code

With sign-in (recommended): run the command, then type “/mcp” in the program and select Wandlix; the browser opens the sign-in.

claude mcp add --transport http wandlix https://wandlix.com/api/v1/mcp

Alternatively with an API key (no browser, for scripts):

claude mcp add --transport http wandlix https://wandlix.com/api/v1/mcp --header "Authorization: Bearer owk_YOUR_KEY"

Cursor

{
  "mcpServers": {
    "wandlix": {
      "url": "https://wandlix.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer owk_YOUR_KEY"
      }
    }
  }
}

VS Code

{
  "servers": {
    "wandlix": {
      "type": "http",
      "url": "https://wandlix.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer owk_YOUR_KEY"
      }
    }
  }
}

Claude Desktop

Claude Desktop connects to remote servers with the small helper mcp-remote (Node.js required). The key is passed through an environment variable.

{
  "mcpServers": {
    "wandlix": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://wandlix.com/api/v1/mcp",
        "--header",
        "Authorization:${WANDLIX_AUTH}"
      ],
      "env": {
        "WANDLIX_AUTH": "Bearer owk_YOUR_KEY"
      }
    }
  }
}

Test the connection

This call lists all tools (insert your key):

curl -X POST https://wandlix.com/api/v1/mcp   -H "Content-Type: application/json"   -H "Accept: application/json, text/event-stream"   -H "Authorization: Bearer owk_YOUR_KEY"   -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Tool calls use the same limits as the HTTP API: size per tool, requests per minute and your daily free quota or credits.

n8n

n8n calls Wandlix with its HTTP Request node. Import one of our ready-made workflows or build the node yourself in three steps.

1. Create the credential

   In n8n create a credential of type “Header Auth”: name Authorization, value Bearer owk_YOUR_KEY.

2. Add an HTTP Request node

   Method POST, URL https://wandlix.com/api/v1/tools/<tool>, authentication “Generic credential type” with the credential from step 1. Send the body as “Form-Data” with the field files of type “n8n Binary File” (field name data) and, if needed, a text field options with the settings as JSON.

3. Use the result

   Set the response format to “File” – the result lands in the binary field data and can go on to Google Drive, e-mail, Dropbox or any other node. Reports such as the e-invoice check come back as JSON.

Ready-made workflows

Download a file and import it into n8n (menu → Import from file). It contains no credentials. Replace the trigger with your own file source.

• Compress PDF (compress-pdf)
• Merge PDF (merge-pdf)
• Word to PDF (word-to-pdf)
• Validate e-invoice (validate-einvoice)
• Make PDF searchable (OCR) (ocr-pdf)
• Convert image (convert-image)

Make

In Make you use the module “HTTP – Make a request”. No special app is needed.

1. Add the module

   Choose HTTP → Make a request. Enter the URL of the tool, method POST.

2. Set the header

   Add the header Authorization with the value Bearer owk_YOUR_KEY. Better: store the key as a Make connection or variable instead of typing it into the module.

3. Set the body

   Body type “Multipart/form-data”. Add a field of type “File” with key files (file name and data come from the previous module, for example Google Drive or e-mail). For tools with settings add a Text field options with JSON.

4. Use the result

   Leave “Parse response” off. The output Data is the finished file; map it into the next module, for example Upload a file.

URL:     https://wandlix.com/api/v1/tools/compress-pdf
Method:  POST
Headers: Authorization: Bearer owk_YOUR_KEY
Body:    multipart/form-data
  files   (File)  -> file name + data from the previous module
  options (Text)  -> {"level":"medium"}

Zapier

Zapier reaches Wandlix with the step “Code by Zapier” (JavaScript). Pass the file URL from the previous step and your key as input data.

// Code by Zapier (JavaScript)
// Input data: fileUrl (file from the previous step), apiKey (your owk_ key)
const source = await fetch(inputData.fileUrl);
const form = new FormData();
form.append("files", new Blob([await source.arrayBuffer()]), "input.pdf");
form.append("options", JSON.stringify({ level: "medium" }));

const res = await fetch("https://wandlix.com/api/v1/tools/compress-pdf", {
  method: "POST",
  headers: { Authorization: "Bearer " + inputData.apiKey },
  body: form,
});
if (!res.ok) throw new Error((await res.json()).error);
const bytes = Buffer.from(await res.arrayBuffer());
return { filename: "compressed.pdf", size: bytes.length, base64: bytes.toString("base64") };

Zapier limits the memory and runtime of code steps, so this suits small and medium files. Larger jobs are better run in n8n, Make or directly over the HTTP API.

Official apps in the marketplaces

The setups above work today with the HTTP API. Ready-made apps in the Zapier, Make and n8n directories need the operator’s own developer accounts and are the next step; until then the templates and instructions on this page are the supported way.

Other tools

Anything that can send a multipart/form-data request works the same way: Power Automate, Pipedream, Node-RED, Apps Script or your own scripts. Use the OpenAPI description at /api/v1/openapi.json to generate clients automatically.