Skip to main content
The Superflow MCP server lets your AI assistant work inside your Superflow workspace. Once it is connected, you can ask in plain words and the assistant reads and works through your site feedback, checks what is broken, runs the AI review agents and hands work to Jira or Slack, from the chat or the editor. Model Context Protocol (MCP) is the open standard these assistants use to talk to outside tools.

What you can ask for

What’s broken on acme-dental.com? Then fix the critical ones.
Which client comments have had no reply from us in 3 days?
Resolve everything tagged copy on the homepage.
Draft a client update for Acme covering what closed this week.
Estimate a run of the Pre-Launch pack on Acme, then run it.
Create a project for northwind-dental.com, invite dana@northwind-dental.com and give me the install snippet.
Post the open critical comments on Acme to #design-feedback.
The assistant acts as you. It sees only what your Superflow account can see, in one workspace, and only with the permissions you approve. Anything that changes data, sends email or spends AI credits shows you a preview and asks first.

At a glance

Pick how to connect

There is one Superflow MCP server. You reach it either directly, by signing in from your AI tool, or through a small package on your computer. Both give the same tools and the same answers, and you can use both.

Sign in from your AI tool

Paste https://mcp.usesuperflow.ai/mcp into your tool, sign in to Superflow and approve the permissions. Nothing to install and no key to copy.Best for claude.ai, ChatGPT, Claude Code, Cursor, Gemini CLI and other remote MCP clients.

Run the package locally

npx -y @usesuperflow/mcp starts a local bridge to the same server. Sign in once with login, or use an API key. Adds a read-only mode.Best for Claude Desktop, CI, a key with narrow permissions, or read-only use.

Set up your AI tool

Claude

claude.ai, Claude Desktop, Claude Code and the Claude API.

ChatGPT

Custom connector on Plus, Pro, Business, Enterprise and Edu.

Cursor

One-click install, or mcp.json.

Gemini CLI

settings.json, or gemini mcp add.

Other MCP clients

Anything that speaks Streamable HTTP.

Troubleshooting

Sign-in, permission and start-up problems.

Find the server URL in Superflow

1

Open Settings, then Integrations

In the Superflow app, go to Settings and open the Integrations tab.
2

Choose MCP

Pick MCP in the side menu.
3

Copy the server URL

Copy the Server URL. The page also has setup steps for claude.ai, Claude Code, ChatGPT, Cursor (with an Add to Cursor button) and Gemini CLI.
When your AI tool connects, Superflow opens in the browser. Sign in (or sign up), check the workspace the connection is for, and look at the permissions it asks for. You can untick any of them. Approve, and the tool loads the Superflow tools it is allowed to use. A connection belongs to one workspace. To use a second workspace, connect again and pick it.
The screen lists only what your AI tool asks for. You can untick permissions there but not add new ones. If your tool asks for no permission in particular, Superflow offers Read workspace, Read projects, Read comments, Read agents, Read analytics, Run agents and Import findings. That is enough to read, report and run reviews, but not to resolve or reply. If you need Edit comments or another write permission and the screen does not list it, create an API key with it, or sign in with the package and name it with login --scope.

Try it without an account

A demo endpoint needs no sign-in. It offers two tools that start a free Superflow review of any public URL and read back the findings:
Add it the same way as the main server. It is a quick way to check that a new client can reach Superflow at all, or to show someone what a review returns before they connect a workspace.