Connect your AI app

1. Sign in to your dashboard. Every account starts on the Free plan.
2. Open Connect in the dashboard, pick your app and choose Create a key for that app. The token (ld_agent_...) is shown once, already filled into the command or config below, so copy it from there.
3. Add it to your app with the command or config for that app (below). Restart the app if it asks.
4. Check it is connected. The Connect page flips to "connected" on your app's first call, and the title bar shows "Agent connected" with the time of the last call. Or ask your app: "Call LaunchDistro's license_status."

The snippets below use <your-token> in place of the token; the Connect page shows them with yours in.

• Claude Code. The command adds the server for every project (--scope user). Use claude mcp list to see it and claude mcp remove launchdistro to take it off. Claude Code can run the command for you: paste it into the chat. Full guide: Claude.
• Claude Desktop. Open Customize, then Connectors, click +, choose Add custom connector and paste the URL. Custom connectors work on Free (one connector), Pro, Max, Team and Enterprise; on Team and Enterprise an owner adds it first under Organization settings, then each member connects. Connectors take a URL and OAuth details, not a custom header, so the token is in the URL (?token=): keep it private, and remove the connector or revoke the token if it leaks.
• Cursor. Merge the entry into ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project) without removing servers already there. Cursor can read the token from an environment variable ("Authorization": "Bearer ${env:LAUNCHDISTRO_TOKEN}") if you prefer not to keep it in the file. Guide: Cursor.
• Codex. The Codex CLI, the IDE extension and the ChatGPT desktop app share ~/.codex/config.toml. To keep the token out of the file, replace http_headers with bearer_token_env_var = "LAUNCHDISTRO_TOKEN" and set that variable. Guide: Codex.
• VS Code. The command adds it to your profile; use it in Agent mode in chat. You can also add it by hand to .vscode/mcp.json or your user mcp.json as a server with "type": "http", a url and a headers entry. Guide: VS Code.
• A browser for the forms. Your agent needs one to fill directory forms. Recommended: your own Chrome through Claude in Chrome, where you are already signed in to directories. Your app's built-in browser works too; it will ask which one to use.

1. "Add my product from this repo to LaunchDistro." It reads the repository and your site, shows you the profile and saves it.
2. "Where does my product qualify?" (match_directories)
3. "Plan the run." (plan_submissions): the order and the reason for each, saved to the dashboard.
4. "Submit to the next 3 directories." It reads each playbook, asks you every decision at once, and fills the forms in your browser.
5. "What have I sent?" (list_submissions), and "check if any are live."

Open your product's folder first. In Claude Code or Cursor, start the agent inside the repository of the product you are launching, so it can read the README, the description and the stack for the profile. In apps with no file access (Claude Desktop, ChatGPT) the agent reads your live site and asks you for what it cannot find.

A badge or link back is added to your repository only if you say you will show one. The agent shows you the change, and committing and deploying stay with you.

The dashboard has a copyable prompt for each of these that already names your product and the next directory. There is also a submit prompt in apps that show MCP prompts.

Each AI app has its own token, so one can be revoked without touching the others. Removing an app from the Connect page revokes its token at once. The plan sets how many apps can be connected: 1 on Free and on a Launch, 3 on Pro. The token is stored only as a hash; if you lose it, remove the app and create a new key.

Any app that connects to remote MCP servers works. Point it at the endpoint with the header:

To see it answer from a terminal, list the tools with one request:

curl -s https://launchdistro.com/api/mcp \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Responses are plain JSON: the server keeps no session between requests, so there is nothing to initialise or keep alive. The tools reference lists every tool's parameters.

Each app documents its own MCP settings; these are the pages the steps above follow.

• Claude Code: connect to MCP servers
• Claude: custom connectors using remote MCP
• Cursor: Model Context Protocol
• Codex: MCP servers
• VS Code: add and manage MCP servers
• MCP specification: transports

• "Unauthorized" or a 401. The token is wrong, was revoked, or lost its Bearer prefix. Create a new key on the Connect page.
• The app says the endpoint is not found. Check the address ends in /api/mcp, with no extra path.
• Connected, but no tools. Restart the app, or start a new chat: many apps read the tool list once per chat.
• The Connect page still says "waiting". It flips on the first tool call, not on adding the server. Ask your app to call license_status.
• "Seats" or app limit reached. Remove an app you no longer use from the Connect page, or move to Pro for 3 apps.
• A tool answers with a limit. Free plan used up, a Pro directory, a Launch that ended or a rate limit: the message names it and the way out. See tool errors.