Classmind

GIF
diagram

Inspiration

Teachers spend hours digging through spreadsheets to understand where students are struggling — time that could go toward actually helping them. We wanted to bring AI-powered insights directly into the tools teachers and students already use, without adding another app to learn.

What it does

ClassMind is an AI teaching assistant for Grade 7 Mathematics that lives inside Slack. Teachers can request exam summaries, analyze individual student performance, and generate variant practice questions — all by @mentioning the bot. Students can describe a concept they're confused about in plain language and get matched learning materials instantly. Any report or practice set can be exported to Google Docs in a single prompt.

How we built it

We built the agent on Elastic Agent Builder with custom ES|QL tools for analytical queries, semantic search across questions, materials, and curriculum, and an MCP-connected Google Docs server for document export. A Python Slack bot bridges Slack events to the agent via the Kibana REST API over Socket Mode.

Challenges we ran into

Designing the agent instructions was the hardest part — LLMs are probabilistic, so the same question can produce different tool choices or response formats on different runs. We had to write precise, structured instructions that constrained the agent's behavior without being so rigid it broke on edge cases. Tuning when to run analytical queries versus semantic search versus export to Docs took significant iteration. Designing permission gating — where teachers see individual student data but students only see class-level stats — required separate bot configurations rather than a single agent.

Accomplishments that we're proud of

We're also proud that teachers can run almost any complicated analytical query — score distributions, per-skill breakdowns, student rankings — simply by asking in natural language, with no SQL or dashboard configuration required. And the fact that it all happens inside Slack — through a channel @mention for shared class-wide insights or a private DM for personal help — means there's no new tool for anyone to learn.