Skip to content

Connect an AI assistant

Connect your AI assistant once and it can work on your organisation’s events whenever you ask. There are two ways in:

  • Sign in with OnStage: claude.ai, Claude Desktop, ChatGPT and Claude Code. You paste the server URL, sign in to your OnStage dashboard in the browser window that opens, and click Allow. Use this if your assistant offers it.
  • Personal access token: every other assistant that supports MCP, such as Cursor, VS Code, Gemini CLI, Codex CLI and Windsurf. You create a token in OnStage and put it in the assistant’s settings. Creating a token needs two-step verification on.

Either way, the server URL is https://api.onstageeventsapp.com/mcp. The AI access page shows it next to Server URL, with a Copy URL button.

Who can do this: Org admins. Other roles see “Only an org admin can manage AI access”.

Where: Account menu (top right) → AI access & two-step

Claude and ChatGPT open your OnStage dashboard in a browser window when you connect. It works the same way for each:

  1. If you aren’t signed in to the dashboard in that browser, sign in as usual with your email and password. If your login is in more than one organisation, choose one when asked.
  2. A page headed Connect Claude to OnStage, with your AI assistant’s name, opens. It shows who you’re Signed in as and the organisation your AI assistant will work in. Each connection works in one organisation. To use another, click Switch organisation and pick it; the page stays open for the new one.
  3. Choose what your AI assistant can do. Read is always included. Draft changes lets it prepare drafts for you to review. If Publish changes is offered, it’s off unless you tick it. Leave it off if you’d rather publish everything yourself from the dashboard.
  4. Click Allow. You’re taken back to your AI assistant, which is now connected. Click Cancel instead if you don’t want to connect it.

You don’t need a code from an authenticator app to connect this way, and you aren’t asked for your password again if you’re already signed in. The page works for 15 minutes after your AI assistant opens it; after that, start again from your AI assistant.

The connection lasts 30 days. After that, your AI assistant asks you to sign in again.

  1. On the AI access page, click Copy URL next to Server URL.
  2. In claude.ai or Claude Desktop, open Settings → Connectors.
  3. Choose Add custom connector.
  4. Name it OnStage and paste the server URL, then select Add.
  5. Select Connect, then sign in with OnStage and click Allow.
  6. Back in Claude, start a chat and ask Claude what it can do in OnStage.
  1. On the AI access page, click Copy URL next to Server URL.
  2. In ChatGPT, open Settings → Apps & Connectors → Advanced settings and turn on Developer mode.
  3. Back in Apps & Connectors, select Create.
  4. Name it OnStage, paste the server URL, choose OAuth for authentication, then select Create.
  5. Sign in with OnStage and click Allow.
  6. Back in ChatGPT, start a chat with OnStage switched on and ask what it can do in OnStage.

ChatGPT can’t use a personal access token. It always signs in with OnStage. If you don’t see Developer mode, your ChatGPT plan or workspace may not allow it.

  1. Run this in a terminal:

    Terminal window
    claude mcp add --transport http onstage https://api.onstageeventsapp.com/mcp
  2. Start Claude Code and type /mcp. Choose onstage. Your browser opens your OnStage dashboard; sign in with OnStage and click Allow.

  3. Ask Claude Code what it can do in OnStage.

Claude Code can use a personal access token instead, if you prefer. Create one, then run:

Terminal window
claude mcp add --transport http onstage https://api.onstageeventsapp.com/mcp --header "Authorization: Bearer <your token>"

Any other AI assistant that supports MCP connects with a personal access token. It sends the token with every request, in a header:

Authorization: Bearer <your token>

Creating a token asks for a code from an authenticator app, so you need two-step verification on. It isn’t asked for when you sign in to the dashboard or connect Claude or ChatGPT, only when you create a token. One authenticator covers every organisation you sign in to.

  1. On the AI access page, find the Two-step verification card and click Turn on.
  2. Enter your Password and click Continue.
  3. Scan the QR code with an authenticator app (Google Authenticator, 1Password, Authy…). If you can’t scan it, type the key shown under “Can’t scan it? Type this key instead:”.
  4. Enter the Six-digit code the app shows and click Turn on.
  5. Save the recovery codes: click Copy codes or Download. Each one works once, in place of a code from your authenticator app, and they won’t be shown again.
  6. Tick I’ve saved my recovery codes and click Done.

The card then shows On and how many recovery codes you have left. Click New recovery codes to replace them; the old ones stop working straight away.

  1. On the AI access page, in Connections and tokens, click New token.
  2. Fill in New AI access token:
    • Name: something you’ll recognise later, such as “Cursor on my laptop”.
    • What it can do: Read is always on. Draft changes is ticked for you. Tick Publish changes only if you want your AI assistant to publish what it drafts.
    • Which events: Every event in your organisation, or Only the events I pick.
    • Expires after: 7, 30 or 90 days.
    • Your password and Code from your authenticator app (or one of your recovery codes).
  3. Click Create token.
  4. Click Copy token. It starts with osm_ and this is the only time it’s shown. Then click I’ve copied it.
  5. Add it to your assistant’s settings as shown below, then ask your assistant what it can do in OnStage.

Treat the token like a password: anyone who has it can act as you, within what you let it do. Keep it in your own user-level settings, as below, never in a project file that’s committed or shared with others. You can’t change a token’s permissions later. Revoke it and make another.

A token limited to particular events can’t see your other events, and it can’t create or copy events.

Add OnStage to ~/.cursor/mcp.json in your home folder:

{
"mcpServers": {
"onstage": {
"url": "https://api.onstageeventsapp.com/mcp",
"headers": {
"Authorization": "Bearer <your token>"
}
}
}
}

Open the Command Palette, run MCP: Open User Configuration, and add OnStage to that mcp.json. VS Code asks for the token the first time it starts the server and keeps it in its own secure storage, so it isn’t written in the file:

{
"inputs": [
{
"type": "promptString",
"id": "onstage-token",
"description": "OnStage token",
"password": true
}
],
"servers": {
"onstage": {
"type": "http",
"url": "https://api.onstageeventsapp.com/mcp",
"headers": {
"Authorization": "Bearer ${input:onstage-token}"
}
}
}
}

Then start onstage, paste your token when asked, and ask Copilot in agent mode what it can do in OnStage.

Add OnStage to ~/.gemini/settings.json. Use httpUrl, not url:

{
"mcpServers": {
"onstage": {
"httpUrl": "https://api.onstageeventsapp.com/mcp",
"headers": {
"Authorization": "Bearer <your token>"
}
}
}
}

Add OnStage to ~/.codex/config.toml. Codex reads the token from an environment variable, so it isn’t written in the file:

[mcp_servers.onstage]
url = "https://api.onstageeventsapp.com/mcp"
bearer_token_env_var = "ONSTAGE_TOKEN"

Then set ONSTAGE_TOKEN in the terminal you start Codex from, or in your shell profile:

Terminal window
export ONSTAGE_TOKEN='<your token>'

Add OnStage to ~/.codeium/windsurf/mcp_config.json. Use serverUrl:

{
"mcpServers": {
"onstage": {
"serverUrl": "https://api.onstageeventsapp.com/mcp",
"headers": {
"Authorization": "Bearer <your token>"
}
}
}
}

Any assistant that supports MCP over Streamable HTTP and lets you set a request header can connect. Give it:

  • URL: https://api.onstageeventsapp.com/mcp
  • Header: Authorization, with the value Bearer followed by your token.

It must use a personal access token. Only claude.ai, Claude Desktop, Claude Code and ChatGPT can sign in with OnStage; any other assistant that tries to sign in is refused.

Who’s told when someone connects an AI assistant

Section titled “Who’s told when someone connects an AI assistant”

When an admin connects an AI assistant or creates a token, your organisation’s admins get a notice in the dashboard notifications (the bell): AI assistant connected or AI assistant given access, with a link to See who has AI access.

Connections and tokens on the AI access page lists the assistants that signed in with OnStage (Connected app) and your personal access tokens (Personal token). Each row shows what it Can do, which Events it reaches, when it was created, when it was Last used and when it expires.

To disconnect one, click Revoke, then Revoke again to confirm. It stops working on the assistant’s next request. Click Show revoked and expired to see old ones and why each one stopped.

Connections and tokens also stop working by themselves when:

  • you change your password, or someone resets it;
  • you’re deactivated or removed from the organisation;
  • you stop being an org admin.

Tokens also stop when you turn off two-step verification.

Turning it off ends every token you have, in every organisation you sign in to; you’ll need to turn it back on to create a new one. Claude and ChatGPT stay connected: connecting them doesn’t need two-step verification.

  1. On the Two-step verification card, click Turn off.
  2. Enter your Password and a code from your authenticator app, or a recovery code.
  3. Click Turn off.