Three ways,
your event lives on Loops

Loops turns your event into a Skill in the builder's IDE, a Co-pilot in the browser, and an AI Judge that explains itself. Here's each one.

Skill

your hackathon, inside the builder's IDE

One command installs your event into Cursor, Claude Code, or any agent on the open .agents standard, aware of your sponsors, bounties, and docs.

Co-pilot

sponsor expertise on tap

Every sponsor's docs become a knowledge graph powering an in-browser agent. Builders ideate against real bounties and self-evaluate before submitting.

AI Judge

evaluation you can interrogate

Every project gets a structured report, what was built, how it fits the bounty, strengths and gaps with code references you can interrogate.

Skill

Your hackathon, inside the builder's IDE

One command installs the event into any AI coding agent that supports the open .agents standard. The agent learns your sponsors, bounties, docs, and judging criteria, no context-switching, no cloud IDE.

npx loopshouse add your-event-slug

✓Works with Cursor, Claude Code, Codex, VSCode, and more.
✓Pulls live bounties and submission requirements into the editor.
✓Builders submit straight from the terminal.

loopshouse — zsh

you@mac ~ % ▋

IDE

your-repo

.claude/skills

loops-spring-ai-hackathon

SKILL.mdadded

.agents/skills

loops-spring-ai-hackathon

SKILL.mdadded

Co-pilot

Sponsor expertise on tap

Every sponsor's docs and starter kits become a knowledge graph powering an in-browser AI agent. Builders ideate against real bounties and self-evaluate before they ship — without leaving the page to read documentation.

✓Grounded in the sponsor's actual docs, not generic answers.
✓Ideate against real bounty criteria.
✓Self-evaluate a draft before submitting.

Co-pilot

streaming.mdauth.mderrors.mdquickstart.mdwebhooks.md

Ingesting sponsor docs…

sdkfeatureconceptservicelibraryother

AI Judge

Evaluation you can interrogate

Every project gets a structured report — what was built, how it fits the bounty, strengths and weaknesses. Then you cross-examine it: ask whether a team actually integrated your SDK, what's genuinely built versus pitched, and get the code references back.

✓Evidence-based reports, not vibes.
✓Ask follow-ups and get cited code in return.
✓Human reviewers confirm the final winners.

AI Judge

Interrogate the AI Judge's evaluation of this project — it reads the repo, checks integrations against the sponsor KG, and cites evidence. Try one:

Why did you give the score you did? Walk me through it.›
Did they actually integrate our SDK / product? Show the code.›
Which bounty criteria are strongest and weakest here, with evidence?›
What's genuinely built vs. just claimed in the pitch?›
Pull up the main entrypoint and explain the architecture.›
Is there anything you couldn't verify that I should check manually?›

For hosts

Run your hackathon on Loops

Hosting your first hackathon? Or your tenth? See the Skill, Co-pilot, and AI Judge run end-to-end against a real event — Nucleate's, Codex Community's, or yours. 30-min walkthrough, no decks.