MCP Integration, a Connection to Your External Systems
Connect your AI Chat to your systems through MCP Integration, and as an AI Agent it actually completes the tasks you define for your visitors - whether that's booking an appointment or anything else.
What Can MCP Integration Do?
With MCP Integration, your AI Chat goes one step beyond answering questions - it can act as a real AI Agent inside the systems you connect it to. Exactly what it can do always depends on which integration you choose and what you've predefined it's allowed to do.
How MCP Integration Works With Your AI Chat
With MCP Integration, your AI Chat doesn't stop at a good answer - if you enable it, it can actually complete the task in a connected system. Connect your Calendly calendar, for example, and a visitor can book a free slot right in the conversation, without ever leaving your site.
The exact capabilities always depend on which system you connect to the AI Chat - the list of available, ready-made integrations keeps growing.
Activating the integration and connecting your account also takes just a few clicks in the admin panel - no developer help or custom coding required.
Secure Authentication for Every Integration
You can choose from four authentication methods: None, for systems that are reachable without any login, or Bearer token, API key, and OAuth, for systems that require authentication. For OAuth, you connect your account to the external service with a single click, and access refreshes automatically afterward.
Every token and secret is stored encrypted, isolated per tenant, so one customer's credentials are never accessible to another.
See Exactly What Actions Your AI Chat Took
Every action is logged in detail: which integration and function the AI Chat called, how long it took, and what the result was - for example, whether it successfully booked an appointment.
This gives you full visibility into how your AI Chat acts on your behalf - and helps you quickly spot an integration that's slow or failing.
AI Chat Permissions
Permissions apply per integration: if your AI Chat gets access to one of your systems, that doesn't automatically open a path to another - each connection runs independently, with its own, separately configured permissions.
Your AI Chat can only carry out the actions you've authorized, and only the way you've described in the system prompt extension - a single switch can revoke access at any time, immediately, with no developer work needed.
MCP (Model Context Protocol) is an open standard that gives an AI model - in this case, your LivePly AI Chat - a unified, secure way to connect to the data and functions of external systems. This lets your AI Chat also work as an AI Agent, meaning it doesn't just answer - it can actually complete a task in a connected system.
It always depends on which system you connect - a calendar app lets it book appointments, for example, while other systems unlock different capabilities as the list of supported integrations grows.
It supports four methods: the first is for systems that are reachable without any login, and the other three are for systems that require authentication - Bearer token, API key, and OAuth. For OAuth, you connect your account with a single click, and access refreshes automatically afterward.
You describe when and under what conditions it should carry out a given integration's action in the system prompt extension - the same way you fine-tune the main system prompt in Behavior Fine-Tuning.
Yes, every action is logged in detail: which integration the AI Chat called, how long it took, and what the result was.
You can choose from LivePly's ready-made integrations, such as Calendly - the list of supported systems keeps growing.
No, activating an integration and connecting your account takes just a few clicks in the admin panel - no coding required.
Yes, every token and secret is stored encrypted, isolated per tenant, so no other customer can access your data or your credentials.
The admin panel shows a status indicator if an integration is offline or failing, and your AI Chat falls back to your configured fallback message or your Knowledge Base content.
Yes, a single switch lets you disable or re-enable an integration's access at any time, without losing its configuration.