Skip to contentLaunchSuite
Create account

The solution

LaunchSuite MCP: your AI operates, you approve with your key

LaunchSuite's MCP server lets Claude or another AI manage your launch, and requires your passkey approval before anything is published or spent. How it works.

LaunchSuite MCP: your AI operates, you approve with your key

LaunchSuite has a remote MCP server with OAuth 2.1 to which you connect your AI assistant, which can read your projects, check your site, plan and leave posts and campaigns prepared. Whatever is public or spends money does not run by itself: it stays pending until a person approves it with a passkey, and every call is logged.

What it does

MCP (Model Context Protocol) is the standard through which an AI assistant connects to a tool and uses it. LaunchSuite's server exposes dozens of tools, grouped by what they cover: projects, website analysis, social presence, site check, directories, UTM links, posts (draft, schedule, publish, cancel and check status), channel connections, media, strategy and content plan, approvals and Apple Ads campaigns.

It connects like any remote MCP server: you add its address in your client, sign in, choose the workspace you give access to and confirm. There are no keys to copy. The server page, with the address and per-client guides, is at Connect your AI (MCP).

The approval gate

It is what sets it apart from handing a key to an AI. The tools fall into three kinds: read tools, which run directly; tools that write internal things, such as a draft or an account's status, which also run; and those that are public or spend money, which do not run, but create an action pending approval.

  • Public: scheduling and publishing a post.
  • Spending: launching or resuming a campaign and changing its budget.
  • Tied to the exact content. A person approves that specific action; if the draft changes afterwards, the approval stops being valid.
  • With a passkey. The approval is signed with your fingerprint or Face ID through a passkey, bound to that content. Anything that spends money is approved only this way.
  • A test enforces it. That no public or spending tool runs without approval is an automated check of the code, not an instruction in a prompt.

What else protects access

  • Access per workspace, with OAuth 2.1: you choose which workspace the AI sees and you can withdraw the permission.
  • Usage limits: 60 calls per minute per connection and 240 per workspace.
  • An audit of every call, without storing the arguments.
  • Your site's text arrives marked as untrusted data, with a fixed warning, so an instruction hidden in a page is not mistaken for an order from you.
  • Spend caps are not raised over MCP: there is no tool to change the monthly cap, which is set only in the panel with your passkey.

How to use it, step by step

  1. Connect your assistant to the MCP server and sign in with your LaunchSuite account.
  2. Choose the workspace you give access to.
  3. Ask for things such as "summarise the state of my projects", "run the site check" or "prepare this week's posts as drafts".
  4. When it proposes to publish or spend, you get a pending action. Open it in the panel, review the exact content and approve it with your passkey.
  5. Check the log to see what your AI did and when.

What it does not do

  • It does not bring its own AI. You use your assistant; the copy and the ideas come from it, not from LaunchSuite.
  • It does not generate images or videos. Your AI creates the pieces with its own tools and you upload them.
  • The passkey is for approving, not for signing in. Access to the panel is still with email and password.
  • It does not remove the risk of an AI making a mistake: it bounds it, because what matters goes through your approval before it happens.

Which plans include it

It is in every plan. Each plan's limits are on the pricing page.

What problem it solves

It solves the dilemma of letting an AI act on your accounts without losing control: neither handing over the keys nor copy-pasting by hand. Applied to money, it is what the Apple Ads module with a spend cap explains, and it is the answer to the fear of ad spend running away. If you want your AI to work on several products at once, combine it with the launch matrix. And to see how it compares with other tools that also offer MCP, there is LaunchSuite vs Postiz and LaunchSuite vs Metricool.

Frequently asked questions

Which AI clients work?

Any compatible with remote MCP servers with OAuth, such as Claude. The connection page explains the steps per client.

Can my AI publish without me knowing?

No. Publishing and spending create a pending action that only runs when a person approves it with their passkey.

What does the AI see of my workspace?

What you allow when connecting: one specific workspace and the server's tools. The text of websites is handed to it marked as untrusted data.

Your whole launch, in one tool

One project per product, each with its own plan, accounts, calendar and directories. See the plans and start the trial.

  • Fear of ad spend running away

    Why a daily budget is not a cap, how spend escapes in a small test and what it takes to advertise without fearing that the bill will spiral out of control.