# Connect an AI client to the Valsight MCP server URL: https://docs.valsight.ai/connect-ai-client-to-valsight-mcp/ Description: Connect your AI client — Claude, ChatGPT, GitHub Copilot, Microsoft Copilot, or Visual Studio Code — to the Valsight MCP server so it can work with live data from your Valsight projects. ## Goal You will have your AI client connected to the [Valsight MCP server](/valsight-mcp-server/), so it can read — and, in a write session, change — data in your Valsight projects on your behalf. Setup takes about five minutes. Every client uses the same server address, the same sign-in, and the same permissions; only the place where you enter the address differs. > **Prerequisites** > > * A Valsight account with administrator permissions > * Network access to your Valsight instance Verify access programmatically before proceeding: `GET {base_url}/api/v1/health` must return HTTP 200 with a valid API key in the `X-API-Key` header. The MCP server must also be **enabled for your Valsight instance** — it is a premium feature, opted into per customer (see [availability and pricing](/valsight-mcp-server/)). The server is remote, so there is nothing to install on the Valsight side; your client connects to it over the internet. ## Supported AI clients | AI client | Where it runs | Setup | | ---------------------- | ------------------------------------------------------------------- | ----------------------------------------- | | **Claude** (Anthropic) | claude.ai, Claude Desktop, Claude Code | [Claude](#claude) | | **ChatGPT** (OpenAI) | chatgpt.com, ChatGPT apps | [ChatGPT](#chatgpt) | | **GitHub Copilot** | Visual Studio Code, GitHub Copilot app | [GitHub Copilot](#github-copilot) | | **Microsoft Copilot** | Copilot Studio agents, published to Microsoft 365 Copilot and Teams | [Microsoft Copilot](#microsoft-copilot) | | **Visual Studio Code** | VS Code agent mode | [Visual Studio Code](#visual-studio-code) | Other clients that support remote MCP servers with OAuth sign-in may also work, but are not tested by Valsight. *Note: AI clients change their settings screens often. The click paths below describe each client at the time of writing; if a menu is named differently in your version, look for the client's MCP, connector, or tool settings and enter the same server address.* ## Your MCP server address Every Valsight instance has its own MCP server address. It names both the MCP server and your instance, so do not copy an address from someone at another company or from an example. From Valsight 7.3.2, copy the address from Valsight: 1. Click your avatar at the bottom of the navigation bar. Your user profile opens. 2. Under **AI Clients (MCP)**, click **Connect an AI Client**. 3. Click the copy button next to **Your server address**. The **Connect an AI Client** page also has checkboxes for the most common [session limits](#restrict-the-session-from-the-start-optional), ready-to-paste setup for Claude Code and Visual Studio Code, and a link to manage your sessions. If your profile has no **AI Clients (MCP)** section, the MCP integration is not enabled for your instance. On 7.3.0 and 7.3.1, ask your Valsight admin or for your instance's address. The address is one continuous string with no spaces — copy it in one piece. In the steps below, `` stands for it. ### Restrict the session from the start (optional) By default a session can **write** — change your models — and see numeric values. To connect with tighter limits, add parameters to the end of the address: with `&` if the address already contains a `?`, otherwise with `?`. | Parameter | Effect | | ---------------------------------- | ----------------------------------------------------------------------------- | | `accessLevel=read_only` | The agent can read but not create, change, or delete anything. | | `dataAccess=restricted` | Numeric values (calculation results, line-item values) are hidden. | | `projects=,` | The session can only touch these project spaces (numeric id or business key). | | `allowProjectDelete=no` | Whole project spaces cannot be deleted. | For example, a read-only session without numeric data adds `accessLevel=read_only&dataAccess=restricted`. From 7.3.2, on the **Connect an AI Client** page, the **Read-only** and **No numeric data** checkboxes add the first two parameters for you. What each limit means is covered in [Session scope and permissions](/valsight-mcp-server/#session-scope-and-permissions). ## Steps Pick your client below, then complete [the sign-in flow](#the-sign-in-flow) when the browser opens. ### Claude #### On claude.ai and Claude Desktop Connectors you add on claude.ai are also available in Claude Desktop. 1. Open the connector settings: your initials (bottom left) → **Settings** → **Connectors**. 2. Click **Add custom connector** and name it `Valsight`. 3. Paste your [MCP server address](#your-mcp-server-address). Leave **Advanced settings** (Client ID / Secret) empty — registration is automatic. 4. Click **Add**, then **Connect**. The login window opens. 5. Once the connector shows **Connected**, enable Valsight from the tools menu (the sliders icon) in any new chat. On Team and Enterprise plans, an owner may need to add the connector for the organization first. #### In Claude Code 1. Sign in to Claude Code (first time only): open a terminal, run `claude`, and sign in with your Anthropic account. 2. Register the MCP server in one command: ```bash claude mcp add --transport http valsight "" ``` Keep the double quotes: they stop the shell from reading the `?` and `&` in the address. 3. Authenticate: run `/mcp` → **valsight** → **Authenticate**. The OAuth flow opens in the browser. 4. Tip: add `--scope user` to the `add` command to make the server available in all your projects. ### ChatGPT Custom MCP connectors in ChatGPT need a paid plan; on Business and Enterprise workspaces an admin may have to allow custom connectors first. 1. Open **Settings** → **Apps & Connectors** → **Advanced settings** and turn on **Developer mode**. 2. Back in **Apps & Connectors**, click **Create** to add a new connector. 3. Name it `Valsight`, paste your [MCP server address](#your-mcp-server-address), and choose **OAuth** as the authentication method. 4. Confirm that you trust the connector, then click **Create**. The login window opens. 5. In a new chat, enable Valsight from the **+** menu (or the tools menu) to use it. ### GitHub Copilot **In Visual Studio Code:** GitHub Copilot uses the MCP servers configured in VS Code. Add Valsight as described under [Visual Studio Code](#visual-studio-code), then open Copilot Chat in **Agent** mode and make sure the Valsight tools are enabled in the tools picker. **In the GitHub Copilot app:** 1. Open the app's MCP server settings and add a new server. 2. Choose a remote (**HTTP**) server, name it `valsight`, and paste your [MCP server address](#your-mcp-server-address). 3. Save. When the app asks you to sign in to the server, allow it — the login window opens. ### Microsoft Copilot Microsoft Copilot connects to MCP servers through **Copilot Studio**: you add Valsight as a tool to a Copilot Studio agent, then publish that agent to Microsoft 365 Copilot or Teams. This is typically done by whoever builds agents in your organization, and your Power Platform admin may need to allow the connection. 1. In Copilot Studio, open your agent (or create one) and go to **Tools** → **Add a tool** → **New tool** → **Model Context Protocol**. 2. Enter a server name (`Valsight`), a short description, and your [MCP server address](#your-mcp-server-address). 3. Under authentication, choose **OAuth 2.0** with **Dynamic discovery**. No client ID or secret is needed. 4. Click **Create**, then create a new connection — the login window opens. 5. Add the tool to the agent, test it in the test pane, then publish the agent to the **Microsoft 365 Copilot** or **Teams** channel. Each user who uses the agent signs in with their own Valsight account, so everyone works within their own permissions. ### Visual Studio Code 1. Open the Command Palette (`Ctrl+Shift+P` / `Cmd+Shift+P`) and run **MCP: Add Server…**. 2. Choose **HTTP**, paste your [MCP server address](#your-mcp-server-address), and name the server `valsight`. 3. Choose whether to add it for your user (all workspaces) or only this workspace. 4. Start the server. When VS Code reports that the server wants to authenticate, click **Allow** — the login window opens. The resulting entry in `mcp.json` (user profile or `.vscode/mcp.json`): ```json { "servers": { "valsight": { "type": "http", "url": "" } } } ``` ### The sign-in flow Sign-in is the same for every client. Afterwards the connection stays active and tokens refresh automatically. 1. The browser opens **id.valsight.com** — Valsight's own identity service. This window is expected. 2. Approve the authorization request ("MCP Client requests access"). This authorizes your client to act with your permissions. 3. Log in with your **existing** Valsight account — your work e-mail and Valsight password, or your company's single sign-on if your instance uses it. Never click **Register**; your account already exists. ## Verify * **In the client:** the Valsight connector or server shows as connected (for Claude Code: `claude mcp list` lists `valsight` as connected). * **End to end:** ask the agent something that needs live project data (for example, to list your projects). A grounded answer confirms the connection works. * **Check the limits:** ask the agent to show the session's access settings. It reports the account, instance, access level, data access, and project scope in force. ## Control the session Scope a session deliberately — by default it can **write**. Besides the [connection parameters](#restrict-the-session-from-the-start-optional), you can change the limits at any time by asking the agent in plain language (for example, "make this session read-only"). **Narrowing** takes effect immediately. **Broadening** — for example from read-only to write — always needs your confirmation: the client shows a confirmation form, or the agent gives you a link to the session's settings page, where you review and save the change. Session scope, access level, and data access, and their defaults, are covered in [Session scope and permissions](/valsight-mcp-server/#session-scope-and-permissions). Your Valsight sign-in stays valid for 12 hours, after which you sign in again. ## Help Login trouble, or projects missing after you connect? Contact your Valsight admin or write to . ## Related * [Valsight MCP integration](/valsight-mcp-server/): what the server is, its capabilities, security, and how session scope keeps a connection contained. * [FAQ: Valsight MCP integration](/frequently-asked-questions-valsight-mcp-integration/): connecting, security, and prerequisites. * [Valsight AI](/ai-assistant/): overview of both ways to use Valsight AI.