# FAQ: Valsight MCP integration
URL: https://docs.valsight.ai/frequently-asked-questions-valsight-mcp-integration/
Description: Frequently asked questions about the Valsight MCP integration — connecting an external AI client to Valsight, security, prerequisites, and what it can do.
Common questions about the **Valsight MCP integration** — connecting your own external AI client to Valsight. This is the technical, external counterpart to the [in-app Assistant](/using-the-ai-assistant/); if you want to chat with the AI inside Valsight instead, see the [FAQ: AI Assistant](/frequently-asked-questions-valsight-ai-assistant/).

**What is the Valsight MCP integration?**

It lets you connect your own external AI client — such as Claude — to Valsight over the Model Context Protocol (MCP), an open standard for connecting AI clients to tools and data. The connected client can then explore and, in write mode, change your models from outside the Valsight app, always within your own permissions.

**How is it different from the in-app AI Assistant?**

The **in-app Assistant** is a conversational assistant you use inside Valsight, aimed at all users. The **MCP integration** lets you drive Valsight from your own external AI client or agent, and is aimed at technical users. Both build on the same model context and both act only within your Valsight permissions.

**Which AI clients can I use?**

Valsight supports:

* **Claude** (Anthropic) — claude.ai, Claude Desktop, and Claude Code
* **ChatGPT** (OpenAI)
* **GitHub Copilot** — in Visual Studio Code and the GitHub Copilot app
* **Microsoft Copilot** — through agents built in Copilot Studio and published to Microsoft 365 Copilot or Teams
* **Visual Studio Code** — agent mode

Step-by-step setup for each is in [Connect an AI client to the Valsight MCP server](/connect-ai-client-to-valsight-mcp/). Other clients that support remote MCP servers with OAuth sign-in may also work, but are not tested by Valsight.

**How do I connect?**

Add your instance's MCP server address to your AI client and sign in when prompted, using your normal Valsight login. There is no single, public endpoint: each instance has its own address. Your instance must also have the MCP integration enabled (see prerequisites). Where to find your address, and the steps for each client, are in [Connect an AI client to the Valsight MCP server](/connect-ai-client-to-valsight-mcp/#your-mcp-server-address).

**What do I need before I can use it?**

* Your Valsight instance must have the MCP integration **enabled**. It is a premium feature, opted into per customer (see "What does it cost?" below).
* A compatible Valsight version — available from Valsight 7.3.0.
* A [supported AI client](/connect-ai-client-to-valsight-mcp/#supported-ai-clients).

**What does it cost?**

The MCP integration is a premium feature, available from Valsight 7.3.0. Through **31 December 2026** it is included for all customers at no additional cost, so you can connect your own AI client, work with your real models, and see what it adds before deciding to keep it. From **January 2027** it becomes a paid add-on. It is opted into per instance and needs some setup, so it is not enabled by default. To turn it on, or to talk through whether it fits your team, write to <support@valsight.com> or contact your Customer Success representative.

**Is it secure?**

You authenticate with your normal Valsight sign-in, and the integration acts only within your permissions — it cannot see or do more than you can. Each connection runs as a session whose scope you control: which projects it can touch, whether it is read-only or write, and whether it can see numeric values. Connections are only accepted for recognized Valsight instances, sign-in only completes for approved AI clients, and every tool call is logged. Details for security reviews are in [Security and data flow](/valsight-mcp-server/#security-and-data-flow).

**Where does my data go?**

The MCP server passes what the agent requests between your Valsight instance and your AI client; it does not run an AI model itself. Content the agent reads is processed by your AI client and its model provider (for example Anthropic, OpenAI, Microsoft, or GitHub) under your organization's agreement with that provider. If numeric values should not reach the AI client, connect with restricted data access.

**Can it change my models?**

Yes, when the session is connected in **write mode**. You control the session's access level (read-only or write) and its data access (restricted or full), as well as which projects are in scope. Connect read-only if you only want it to explore.

**How do I control what a session can do?**

Session scope can be fixed when you connect, by adding parameters such as `accessLevel=read_only` or `dataAccess=restricted` to the server address, and you can change it later by asking the agent in plain language. Broadening access always needs your confirmation. It covers the target instance, the projects in scope, the access level, and data access. Two further options are separate: changing access rights is off by default, and deleting whole projects is allowed unless you block it.

**How long does a connection last?**

A session ends after a period of inactivity, when access is revoked, or when an administrator ends it. Your Valsight sign-in stays valid for 12 hours, after which you sign in again.

## Related

* [Valsight AI](/ai-assistant/): overview of both ways to use Valsight AI
* [Valsight MCP integration](/valsight-mcp-server/): what it is, capabilities, and security
* [Connect an AI client to the Valsight MCP server](/connect-ai-client-to-valsight-mcp/): supported clients and step-by-step setup
* [FAQ: AI Assistant](/frequently-asked-questions-valsight-ai-assistant/): questions about the in-app Assistant