Kadens runs a Model Context Protocol (MCP) server. Connect your AI assistant once, then ask it in plain words to work with your account: events, orders, students, schedule and website.
A few things people ask once their assistant is connected.
One address for every client. It speaks MCP over streamable HTTP.
Pick your client. They all reach the same server, with the same permissions and the same safety rules.
Claude on the web, desktop and mobile.
ChatGPT on the web. Your account and workspace must allow developer mode and custom MCP connections.
The terminal agent. Run this once inside your project.
After adding the server, start Claude Code in the same project, run /mcp and select kadens to authenticate. Follow the browser approval flow. Kadens will describe Claude Code as software on your computer.
Add this to your Cursor MCP settings and replace the placeholder with an API key.
Any client that supports remote MCP servers over HTTP works. Paste the endpoint above. If the client supports OAuth, it signs in through Kadens. If not, send an API key in this header.
Authorization: Bearer <YOUR_API_KEY>Claude, ChatGPT and Claude Code sign in with OAuth 2.1. Kadens opens a page where you sign in, choose the profile the assistant acts as and approve the permissions it asks for. Nothing is shared until you approve.
Each connection belongs to one profile. To work with another studio or profile, connect again and choose it.
Permissions are grouped into scopes, listed with each group of tools below. Untick any you do not want to grant: the assistant can only see and do what those scopes allow.
Clients without OAuth, such as Cursor, use an API key you issue in your dashboard under Integrations, then MCP. You choose its scopes when you create it and can revoke it at any time.
Issue an API keyThe tools the server offers right now, grouped by area and read live from the server. Each group lists the scopes it needs. Tool names and descriptions are in English, exactly as your assistant reads them.
Beyond the tools above, your assistant can run any of the business actions listed below through five general tools.
kadens_actions_search finds actions by the words you would use to describe them.kadens_action_describe shows the input and output schema of one action.kadens_action_read runs an action that only reads.kadens_action_run runs an action that changes data.kadens_action_run_destructive runs a change that cannot be undone. Your assistant app asks you to confirm it first.actions.delegate permission plus the permission of each action’s area, for example events.write.kadens_actions_search with create event.kadens_action_describe with event.event.create to read the input schema.kadens_action_run with event.event.create, passing a body built from that schema.Every business action an assistant can run through the action tools, grouped by area and read live from the server. Ids and summaries are in English, exactly as your assistant reads them.
Kadens never receives your conversation with the assistant, only the tool calls it makes. We log the tool name, the outcome, the time and the credential used, never the content of what the assistant sends. What the assistant's provider does with the data it receives is governed by that provider's own privacy policy.
Read the privacy policyA connection that fails, a tool that does not behave as described, or a question before you connect? Write to us and tell us which client you use.
Contact usWas this helpful?