A: Hey Jos,

Good questions about using AgenticFlow with protected sites!

1. Embedding a Chatbot on an Internal/Password-Protected Site:

Yes, you can embed the AgenticFlow agent chatbot widget on an internal WordPress site or a client dashboard that requires login. The embedding script itself is just JavaScript. As long as the logged-in user's browser can load scripts from agenticflow.ai, the chat widget will appear and function on the page.

The key thing to remember is that the agent's knowledge comes from what you've configured within AgenticFlow (uploaded documents, FAQs, website crawls of public URLs). The agent cannot automatically scrape or learn from the content behind the password/IP protection of your internal site just by being embedded there. Its knowledge base needs to be built separately.

2. Accessing Data from Protected Sites/Dashboards:

If you need the agent or a workflow to fetch data or perform actions on a site that requires a login or is IP-restricted, standard web scraping won't work. You have two main approaches:

Using the API Call Node (if an API exists): If your internal site or the client dashboard has an API, you can use the API Call node within an AgenticFlow workflow. You would need to configure this node to handle the necessary authentication (e.g., sending username/password, API keys, or tokens in the request headers or body, according to that site's API documentation).
Building a Custom MCP Server (Recommended for internal/complex auth): A more robust and secure method is to create your own small MCP server application. This server could run within your network (if IP protected) or be configured with the necessary credentials to log into the site/dashboard. It would expose simple actions (like "Get Client Data") to AgenticFlow. Your AgenticFlow agent/workflow then securely calls your custom MCP server, which handles the authentication and fetches the data from the protected source. You can find guidance on building MCP servers here: https://modelcontextprotocol.io/quickstart/server
So, embedding the chat interface is easy, but accessing data from behind the login requires using APIs via the API Call node or setting up a custom MCP server to handle the authentication.