Chat tools and AI assistants · Coming soon

MCP

The Model Context Protocol is the open standard that lets an AI assistant call out to external tools: the assistant is the client, and whatever it calls is the server. Claude, ChatGPT and developer tools such as VS Code and Cursor already speak it as clients. Lectora's own assistant uses MCP the same way today, as a client, reaching an in-process Canvas tool server when it reads a course. Turning Lectora into something Claude or ChatGPT could call is the other direction entirely: Lectora would have to build and host its own MCP server, decide exactly which tools it exposes, and put institution-level authorization in front of it. None of that exists yet.

Get in touch about MCP

Not yet verified

  • There is no Lectora MCP server. Nothing can be added to Claude, ChatGPT or any other MCP client today.
  • Which tools it would expose, and how an institution would authorize and revoke access, is not designed yet.
  • Even if a tool list existed, nothing decides yet which of Lectora's actions could safely run without the same human approval every other Lectora workflow requires.

How it connects

In MCP terms a 'tool' is a named, schema-described action a server exposes and a client calls with structured arguments, the same shape whether the server is a calculator, a database, or Lectora. If Lectora hosted one, Claude or ChatGPT would discover its tool list on connecting and could, in principle, call something as simple as a status check on a grading batch, mid-conversation. Today Lectora sits on the other side of exactly this exchange: its own assistant is an MCP client that calls a Canvas tool server running in the same process, to read submissions and rubrics, never the reverse. Nothing about the outward-facing server, its tool list, or its authorization model is designed yet.

Data and access

Lectora reads only what the assessment needs, and writes nothing back until an educator approves it. The Canvas access page documents the full scope, and our privacy policy, subprocessors and DPA cover the legal side.

Common questions

No. The built-in assistant runs inside Lectora and is itself an MCP client, reaching an in-process Canvas tool server when it reads a course. This entry is the other direction: Lectora offered as a server that an outside assistant, such as Claude or ChatGPT, could call. That direction is not built.

Related integrations