Cademí MCP
Cademí MCP is a remote Model Context Protocol server that lets AI assistants work with your Cademí account. You connect it once, authorize it by signing in to Cademí, and the assistant can answer questions and make changes on the Cademí API v3 with the permissions of the administrator who authorized it.
The server address is:
https://mcp.cademi.dev/mcpIt uses the Streamable HTTP transport and OAuth. You never copy an API key: the connection is authorized on Cademí itself, is bound to one account, and acts as the administrator who approved it.
What it can do
- Composite tools for common questions. They answer a whole support question in one call, such as why a student cannot access a course or what the full structure of a course is. They accept students by e-mail, name, or public ID and products by name or public ID, and ask back when a name matches more than one.
- One tool per API operation. Every operation of API release 3.19.0 is a tool except four: the two that return a credential secret, the event stream itself, and the OpenAPI document. Each tool has the permission, input, and output of the operation it calls. For the parameters and fields of each operation, see the API reference.
- Context for the assistant. The Cademí vocabulary and the account the connection works on are available to the assistant as context, and ready-made investigations cover an access problem, a student's experience, and the structure of a course.
Tools that change data are annotated, so clients that support approval prompts ask you before they run them. The assistant sees the description, input, and permission of every tool when it connects.
Quickstart
- Connect. Add
https://mcp.cademi.dev/mcpas a remote MCP server in your client. In Claude, add it as a custom connector; in Claude Code, runclaude mcp add. See Connect a client. - Authorize. The client opens the Cademí authorization page. Choose your account, sign in as an administrator, and approve the connection. See Authorization and permissions.
- Ask. Ask about your account in plain language. The assistant chooses the tools, and asks for your approval before a tool changes data, when the client supports it.
Example questions
- "Why can't [email protected] access the Marketing course?"
- "What is João Silva enrolled in, and which products can he not access?"
- "Show me the full structure of the Marketing course. Which lessons are still drafts?"
- "How is the Marketing course sold, and which access schedules does it have?"
- "Which account am I connected to, and how many API requests did it make in the last 24 hours?"
- "Rename the course Marketing 101 to Marketing Essentials."
The first five are read-only. The last one changes data, so the assistant asks for your approval first when the client supports it.
MCP, CLI, or API
| Use | When |
|---|---|
| Cademí MCP | You want an AI assistant to investigate and act on your account in a conversation, and approve each change in your client. |
| Cademí CLI | You work in a terminal or in scripts and CI, forward events to a local application, or apply declarative configuration. |
| Cademí API v3 | You build your own integration: your code calls the API directly and receives webhooks. |
All three use the same API v3, which checks the permissions of every call. The complete list of operations is in the API reference.
Report a problem
- Bugs in the MCP server: open an MCP problem report. Include the client, the tool name, the error
code, and therequest_idwhen the error has one (Errors and limits). - Feature requests: open a feature request.
- Questions about using the server: Discussions.
- Security vulnerabilities: report them privately at github.com/minhacademi/developers/security, never in an issue.
- Questions about your account, plan, or data: contact Cademí support.
Issues and discussions are public. Never paste an API key, a token, a conversation with personal data, or data about your users.