Getting Started

How it works

Every site you host lives inside a project. A project holds your access settings - a password or SSO rules - and a history of deployments. Each deployment is an immutable snapshot of the files you upload; the most recent one is the version your viewers see. Nothing is ever exposed to search engines or the public web.

You can create and manage all of that from the dashboard. To let an AI assistant do it for you, connect the MCP server (which gives the assistant the tools) and install the deploy Skill (which teaches it how to use them).

1. Create an account

Sign up and create (or join) an organization. Everything you host - projects, deployments, and members - belongs to that organization.

Create your account →

2. Create a project

From the dashboard, click New project and give it a name. Then choose how it is protected - every project is gated, and it can never be open to the public internet:

• Password - visitors enter a password you set. LaunchCloud can generate a strong one for you.
• SSO - visitors sign in through one of your organization's configured auth providers (Google or SAML). Configure providers under your organization's settings first, then pick one here.

A project uses one form of auth at a time. Switching from a password to a provider (or back) replaces the old setting, and you can change it whenever you like from the project's settings without redeploying.

Once created, the project page shows its project id and its URL, https://<projectId>.launchcloud.ai. Upload a folder of static files there to publish your first deployment, or let your assistant do it - that's next.

3. Connect the MCP server

LaunchCloud exposes its project API as a Model Context Protocol server, so assistants like Claude, Cursor, and VS Code can list, create, and manage your projects and deployments on your behalf. The server lives at a single endpoint:

https://launchcloud.ai/mcp

It speaks the streamable HTTP transport and is protected by OAuth. The first time a client connects, it opens a browser so you can sign in to LaunchCloud and authorize the connection. Once authorized, every tool call runs against the organization you signed in with.

For clients that can't run a browser flow - CI, scripts, headless agents - send a personal API key instead, as an Authorization: Bearer header. Create one under Account → API keys in the dashboard.

To revoke a client's access, sign out of that session from your organization's settings, or delete the API key it uses.

Claude Code

Add the server from your terminal:

claude mcp add --transport http launchcloud https://launchcloud.ai/mcp

The next time a tool from the server is used, Claude Code prompts you to authenticate in the browser. Run claude mcp list to confirm it's connected, or /mcp inside a session to check status and re-authenticate.

Claude Desktop and claude.ai

Open Settings → Connectors → Add custom connector, then enter:

• Name: LaunchCloud
• URL: https://launchcloud.ai/mcp

Save the connector and complete the browser sign-in when prompted. The tools appear under the connector once authorized.

Cursor

Add the server to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project):

{
  "mcpServers": {
    "launchcloud": {
      "url": "https://launchcloud.ai/mcp"
    }
  }
}

Reload Cursor, then open Settings → MCP and click Login next to the LaunchCloud server to authorize it.

VS Code

Add the server to .vscode/mcp.json in your workspace:

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

Run the MCP: List Servers command, start the LaunchCloud server, and complete the browser sign-in when VS Code prompts you.

Other clients

Any MCP client that supports the streamable HTTP transport with OAuth will work. Point it at https://launchcloud.ai/mcp and let it drive the sign-in flow - no manual token handling is required.

Available tools

Every tool acts on your authenticated organization.

Deploying is a two-step process. create_deployment takes a projectId and a paths array - just the relative paths you are about to deploy, like index.html or assets/app.css. It returns the new deployment along with one signed PUT URL per path. Your assistant then uploads each file's bytes directly to its URL.

File contents never travel through the MCP server, so the size of a deployment isn't limited by how much your assistant can write in a single message. A 263 KB report and a two-line HTML page take the same one round trip.

Each upload is a plain PUT of the file's bytes with the headers returned alongside its URL - no signing or hashing on your side. The signed URLs expire an hour after they're issued.

Only static web content is accepted; unsupported file types are rejected before any URL is signed. A deployment goes live the moment it's created, so anything not yet uploaded returns a 404 until it lands - upload index.html last. Deployments of more than 200 files are uploaded in batches by calling create_deployment again with the deploymentId it returned.

4. Install the deploy Skill

The MCP server gives your assistant the tools. The deploy Skill gives it the know-how: it loads automatically whenever you ask to deploy, publish, or host something on LaunchCloud, and walks the assistant through resolving or creating a project, packaging files with the right encodings, choosing access controls, and shipping updates.

You only need it once per account or organization. Connect the MCP server first

• the Skill drives those tools.

The Skill is a single SKILL.md file:

Download SKILL.md

Place it inside a folder named deploy-with-launchcloud so the structure looks like this:

deploy-with-launchcloud/
└── SKILL.md

Install on claude.ai

1. Open Settings → Capabilities → Skills.
2. Click Upload Skill and select the deploy-with-launchcloud folder (zip it first if your browser requires an archive).
3. To share it with your whole team, an organization owner uploads it under the organization's Skills instead of a personal account.

The Skill activates automatically the next time you ask Claude to deploy something to LaunchCloud.

Install in Claude Code

Drop the folder into one of the Skills directories:

# Available in every project on this machine
mkdir -p ~/.claude/skills
cp -r deploy-with-launchcloud ~/.claude/skills/

# Or scope it to a single project
mkdir -p .claude/skills
cp -r deploy-with-launchcloud .claude/skills/

Run /skills inside a session to confirm it's listed.

Use with the Claude API

Include the Skill when you create your agent or session per the Agent Skills guide, pointing at the deploy-with-launchcloud folder. The model loads it on demand when a request matches.

Other assistants

Any assistant that supports the Agent Skills (SKILL.md) format can use the same folder - install it wherever that assistant reads Skills from.

5. Deploy from your assistant

With the server connected and the Skill installed, ask for a deploy in plain language:

Deploy this HTML as a new LaunchCloud project called Release notes and protect it with a password.

The assistant creates the project, uploads the files as a deployment, and reports the live URL along with any password it generated. If it doesn't, check that the MCP server is still authorized (/mcp in Claude Code, or the connector's status in claude.ai).

Other things you can ask for once it's set up:

• "List my LaunchCloud projects."
• "Deploy this folder to my Marketing site project."
• "Show the most recent deployments for Release notes."
• "Switch the Docs project to Google SSO instead of a password."

6. Share access

Send the link along with the password, or require SSO locked to your company's domain so teammates get in automatically. You can rotate the password, switch auth providers, or revoke access at any time from the project's settings - the change takes effect immediately, with no redeploy.