HackClaw

AI-powered Discord assistant built for the Gies AI for Impact Challenge—a 24-hour hackathon where students build AI agents with no-code and low-code tools. HackClaw answers questions, routes help tickets to mentors, and learns from every resolution.

Project Lead: Ash Castelino

24h Hackathon Coverage

2 Channels (Discord + Email)

KB Self-Building Knowledge

View on GitHub

What HackClaw Does

Three systems working together to support participants and mentors

🧠

Brain-Routed Knowledge

Topic-specific brain files—schedule, rules, tracks, judging
Router reads the index to pick the right section
Loads only targeted content for accurate answers
No context bloat, no hallucinated rules

🎟️

Help Ticket Queue

/helpme posts to #help-queue with mentor buttons
Claim / Unclaim / Resolve directly from Discord
Mentors auto-added to team channels on claim
Unclaimed tickets re-ping after 10 minutes

💡

Learning Loop

New tickets check past resolutions via embedding similarity
Suggests prior solutions before creating a ticket
Resolving a ticket prompts “how did you fix it?”
Answers grow the knowledge base automatically

🎓

Copilot Studio Expert

Dedicated #ask-hackclaw-copilot channel
Answers only from the Copilot Studio textbook
No general hackathon knowledge leaks in
Keeps tool-specific help focused and reliable

Why Build It

Hackathons don’t have enough mentors, and the same questions repeat all night

Answer the FAQ Instantly

Most hackathon questions have been asked before. HackClaw answers schedule, rules, and tooling questions in seconds—so mentors can spend their time on real technical problems.

A Queue, Not a Flood

Without a system, help requests get lost in channel noise. The ticket queue gives mentors a clear list, with claim buttons and reminders so nothing falls through.

Gets Smarter as the Night Goes On

Every resolved ticket adds to the solutions KB. Common issues get auto-suggested answers on the second occurrence—so by 3am, HackClaw is handling what took mentors all evening.