CourseAdvisorAI

Why this exists

Most course-planning tools are either a static catalog browser (no intelligence) or a chatbot bolted on top (no real access to app state). CoAdvisor.AI instead exposes its core actions as WebMCP tools — real functions, backed by a real database, that an agent visiting the page can call directly. The agent isn't guessing its way through the UI; it's using the same typed, documented tools a developer would.

Two real decision points this is designed around:

• Prospective students comparing whole degree programs before applying (compare_paths)
• Upperclassmen choosing between specializations/tracks, and checking whether their transcript actually supports a given path (check_feasibility)

How human + agent collaboration works here

Every tool is a plain JavaScript function. The human-facing UI (buttons, dropdowns) and the WebMCP registerTool() execute() callbacks both call the exact same functions — there's no separate "agent path" and "human path." Whoever acts, the result is the same real write to MongoDB, visible to the other party on their next look at the page (or reload).

Agent actions are visually and structurally distinct from human ones:

• Advisor notes and conflict flags render as agent-authored callouts, separate from plan entries
• The Session Log tags every entry human or agent, in order, so the full back-and-forth is auditable

Built With

• javascript
• mongodb
• python
• webmcp