Connect your own agent

Add Vistazo to Claude, ChatGPT, Claude Code, Cursor, Codex, Gemini CLI or VS Code.

1. Add Vistazo to your client
2. Sign in and pick your workspaces
3. Make your first calls

1. Add Vistazo to your client

Vistazo is one address, https://mcp.vistazo.ai/mcp, added wherever your client lists remote servers. Clients register themselves, so there is no key to create, and most open a browser to sign you in on the first call. Settings → Agents in the app has the same setup for each client, ready to copy.

Two more addresses suit clients that handle many tools differently. All three reach the same tools with the same sign-in, and differ only in what your client lists up front.

Address	Lists	For
`https://mcp.vistazo.ai/mcp`	Every Vistazo tool, plus four that find and run your sources' tools	Most clients
`https://mcp.vistazo.ai/mcp/slim`	The ten tools a dashboard build needs, plus the same four	Gemini CLI, VS Code, Windsurf, or any client that loads every tool up front
`https://mcp.vistazo.ai/mcp/full`	Every Vistazo tool and every source tool	Claude Code, Codex, Cursor, or any client that loads tools only when it needs them

If you turn off your client's own tool search, or it runs a model without one, switch from the full address to the slim one.

Your connected sources covers the four tools that reach the rest.

Claude (web and desktop app)

A custom connector needs no config file, and it carries over between claude.ai and the desktop app.

In Claude, open Customize → Connectors, click + Add, then Add custom connector.
Name it Vistazo and enter https://mcp.vistazo.ai/mcp as the server URL.
Under OAuth client, leave Use Claude's published identity selected, then click Add.
Sign in to Vistazo when prompted.

On a Claude Team or Enterprise plan, an owner first adds the connector under Organization settings → Connectors. Members then click Connect beside it.

ChatGPT

Open chatgpt.com/plugins, click +, then Add custom MCP server.
Name it Vistazo and enter https://mcp.vistazo.ai/mcp under Connection. Leave OAuth as the authentication.
Accept the risk notice, then click Create as a plugin.
Sign in to Vistazo when prompted. In a new conversation, type @Vistazo to use it.

Under the plugin's Permission setting, Allow low-risk tools lets reads run without a prompt. Tools that change something still ask before each change.

Claude Code

Run this in a terminal. Add --scope user to register it for all your projects. Claude Code loads tools only when it needs them, so it takes the full address.

claude mcp add --transport http vistazo https://mcp.vistazo.ai/mcp/full

Cursor

Add this to ~/.cursor/mcp.json, or to .cursor/mcp.json in a project. Cursor asks you to sign in. Cursor loads tools only when it needs them, so it takes the full address.

{
  "mcpServers": {
    "vistazo": { "url": "https://mcp.vistazo.ai/mcp/full" }
  }
}

Codex

Run both commands in a terminal. Codex does not sign in on the first call, so the login command opens the browser. Codex loads tools only when it needs them, so it takes the full address.

codex mcp add vistazo --url https://mcp.vistazo.ai/mcp/full
codex mcp login vistazo

Claude Desktop (config file)

Use this for a Desktop build without custom connectors. Add it to claude_desktop_config.json, found under Settings → Developer in Claude Desktop, then restart the app.

{
  "mcpServers": {
    "vistazo": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.vistazo.ai/mcp"]
    }
  }
}

VS Code

Add this to .vscode/mcp.json, or to your user mcp.json. VS Code asks you to sign in. VS Code loads every tool up front on some models and stops at 128, its own included, so it takes the slim address.

{
  "servers": {
    "vistazo": { "type": "http", "url": "https://mcp.vistazo.ai/mcp/slim" }
  }
}

Gemini CLI

Run this in a terminal. Gemini CLI opens a browser to sign you in on first use. Gemini CLI loads every tool up front, so it takes the slim address.

gemini mcp add --transport http vistazo https://mcp.vistazo.ai/mcp/slim

Any other client

Add the address wherever your client lists remote servers. The client needs to support streamable HTTP with OAuth 2.1.

Signing in opens a consent screen that asks which workspaces this connection may reach. If you belong to one workspace, the screen skips the question and connects that one. If you belong to several, tick the ones this connection may use. You need at least one.

The screen also says where your answer goes. An app that publishes its own identity shows its web address beside its name, and an app that redirects somewhere else shows that address too. Desktop apps such as Claude Code receive the answer on your own computer, so nothing extra is shown. Only approve a connection you started yourself.

To change the set later, disconnect Vistazo in your client and add it again. To cut a client off, revoke it under Settings → Security → Authorized apps (Security).

3. Make your first calls

Every tool that works in a workspace takes a workspaceId, and the server never assumes one (Which workspace does a call run in?).

get_me lists your workspaces. Pick one that is reachable.
list_dashboards shows what exists in it.
get_dashboard reads one dashboard, its written summary included.

A first build walks through a whole dashboard, tool by tool. The tools themselves are in Widgets. If the workspace has Skills, have your agent read the one that covers its task first.

Contents