Skip to content

Kuali Connector Preview

Connect Kuali to your AI assistant.

Ask Claude, Copilot, or any AI assistant that supports local MCP servers to pull data, build apps, move workflows, and run reports against your Kuali instance — in plain English, right from the chat you already use.

Get started in 5 minutes Prompt library

Natural-language operations

Ask your assistant to list stalled documents, build a new app from a PDF form, or import a spreadsheet of users — and it will.

Works with every major AI client

Claude Desktop, Claude Code, Codex CLI, Gemini CLI, GitHub Copilot, VS Code — one command wires the Connector into whichever you use.

Your data stays yours

API keys live in your OS keychain. The assistant calls your Kuali instance directly from your machine — nothing routes through a third-party server.

Pick your path

  • Getting started

    Install the Connector, connect your Kuali instance, and wire up your first AI assistant in five minutes.

    Quick start

  • Prompt library

    Copy-paste prompts for build apps, Curriculum Management, Sponsored Programs, CSV imports, workflow analysis, and chart reports.

    Prompts

  • Installation

    Install via Homebrew, npx, a one-line installer, or a direct binary download. macOS, Windows, and Linux supported.

    Install

  • AI assistants

    Set up Claude Desktop, Claude Code, Codex, Gemini, Copilot, or VS Code. Pick read-only mode if you only want the assistant to look.

    Guides

  • Command reference

    The complete kuali CLI — every resource, subcommand, and flag, with examples for both humans and scripts.

    Reference

  • MCP tool reference

    Every MCP tool your AI assistant can call — 91 resource tools plus 3 connection-management tools (94 total) — grouped by resource, with read-only and destructive markings.

    MCP tools

What is the Kuali Connector?

The Kuali Connector is a small program that runs on your computer and lets your AI assistant talk to your Kuali instance on your behalf. It's a local MCP server — it speaks the Model Context Protocol (MCP) over stdio from your own machine, which is how tools like Claude Desktop, Claude Code, Codex CLI, Gemini CLI, and GitHub Copilot plug in external capabilities.

Once installed, you can ask your assistant things like "create a Travel Authorization app with a PDF I just uploaded," "import these 2,400 rows into the Human Ethics submissions dataset," or "find proposals that haven't moved in two weeks and draft a nudge." The assistant will call the Connector's tools, run the right GraphQL and REST calls against your Kuali instance, and stream the results back into the conversation.

The same binary also works as a full-featured CLI (kuali apps list, kuali documents create, kuali export csv, …) — handy for scripts, CI pipelines, or the moments when you'd rather just type a command.

Is this a good fit?

The Connector is for people who already interact with Kuali regularly — curriculum coordinators, research administrators, sponsored-programs officers, build-app owners, and platform admins — and who want their AI assistant to do the routine work for them. If your team uses Claude, Copilot, Gemini, or any other AI client that supports local MCP servers, you can put Kuali in front of it with one command. (ChatGPT isn't supported yet — see the FAQ for why.)

If you only log into Kuali occasionally, the web app is probably still the fastest path.

Is it secure?

Using the Connector doesn't expand what anyone can see or do in Kuali. Here's what that means in practice — one thing to read carefully, and five guarantees that hold by default.

Read this one

Data you ask about is sent to your AI provider

That's how AI tool use works: when the assistant calls a Connector tool, the result comes back to it for reasoning, and the content of that response is sent to your AI vendor's model (Anthropic, OpenAI, Google, GitHub, …) along with the rest of the conversation. Your API key isn't shared — but the data the tool returns becomes part of the prompt.

Follow your institution's policy on what may be shared with third-party AI services. If your campus restricts FERPA-covered records, HIPAA data, export-controlled research, or sponsor-confidential material, that restriction applies here. Check with your IT, compliance, or research-office contact before pointing the Connector at datasets you're unsure about — and consider read-only mode with a low-privilege API key to narrow what the assistant can ever pull.

For a deeper look: Read-only mode, how data flows through the assistant, reporting a security issue.