Thank you for being part of the Built with Opus 4.7: a Claude Code hackathon as a builder 👋 This document contains all the key details you’ll need to make the most of the event.
Join the hackathon Discord server here: https://anthropic.com/discord
Please note: We will assign a custom role to your discord handle so you will be able to see the hackathon specific channels.
-
Tuesday, April 21st
-
12:00 PM (EST): Virtual Kickoff – rules, prizes, judging, technical talks
-
12:30 PM (EST): Hacking officially begins; team formation on Discord
-
5:00–6:00 PM (EST): Anthropic office hours (#office-hours)
-
-
Wednesday, April 22nd
-
12:00–1:00 PM (EST): Live Session One - AMA with Thariq Shihipar, Member of Technical Staff at Anthropic (Claude Code)
-
5:00–6:00 PM (EST): Anthropic office hours (#office-hours)
-
-
Thursday, April 23rd
-
All Day: Hacking continues
-
11:00 AM–12:00 PM (EST): Live Session Two - Overview of how to use Claude Managed Agents with Michael Cohen, Member of the Technical Staff at Claude Labs
-
5:00–6:00 PM (EST): Anthropic office hours (#office-hours)
-
-
Friday, April 24th
-
12:00–1:00 PM (EST): Live Session Three - with Mike Brown, 1st place winner of Built with Opus 4.6: insights, learnings, and where is he now?
-
5:00–6:00 PM (EST): Anthropic office hours (#office-hours)
-
-
Saturday, April 25th
-
All Day: Hacking continues
-
5:00–6:00 PM (EST): Anthropic office hours (#office-hours)
-
-
Sunday, April 26th
-
12:00–1:00 PM (EST): Live Session Four - Michal Nedoszytko, 3rd place winner of Built with Opus 4.6: insights, learnings, and where is he now?
-
5:00–6:00 PM (EST): Anthropic office hours (#office-hours)
-
8:00 PM (EST): Submissions due via CV platform
-
-
Monday, April 27th
- All Day: First Round Judging (asynchronous)
-
Tuesday, April 28th
-
12:00 PM (EST): Final Round Judging
-
Top 6 teams announced (#announcements)
-
Final round Judging
-
-
12:45 PM (EST): Closing Ceremony
- Top 3 revealed, celebration & closing remarks
-
To keep things fair and aligned with our goals, all teams must follow these rules:
-
Open Source: Everything shown in the demo must be fully open source. This includes every component - backend, frontend, models, and any other parts of the project - published under an approved open source license.
-
New Work Only: All projects must be started from scratch during the hackathon with no previous work.
-
Team Size: Teams may have up to 2 members.
-
Banned Projects: Projects will be disqualified if they: violate legal, ethical, or platform policies, use code, data, or assets you do not have the rights to.
Start from a real problem in a real place: your work, your community, a field you're close to. The process that takes weeks and should take hours. The thing someone you know still does by hand. Domain expertise beats credentials — show us the thing only you'd know to build.
Looks like: the process that takes weeks and should take hours. The decision was made on gut because the data's too scattered to use. The thing someone you know still does by hand.
Start from something that doesn't exist yet: a new way to work, learn, or make that only makes sense now that the tools have changed. An interface without a name. A workflow from a few years out. The best projects here are easier to demo than to explain
Looks like: an interface that doesn't have a name yet. A workflow that changes how you do the thing, not just how fast. A first draft of how this will work in a few years when Claude is even more capable.
Claude Managed Agents: get to production 10x faster (blog)
Scaling Managed Agents: Decoupling the brain from the hands (blog)
Building Agents with the Claude Agent SDK
Building multi-agent systems: when and how to use them
Best practices for prompt engineering
Extending Claude’s capabilities with skills and MCP servers
Skills explained: How Skills compares to prompts, Projects, MCP, and subagents
Building agents with Skills: Equipping agents for specialized work
Claude Code power user customization: How to configure hooks
Claude Quickstarts - A collection of projects designed to help developers quickly get started with building deployable applications using the Claude API
A Complete Guide to Building Skills for Claude (eBook)
Claude Cookbooks and Cookbooks Github Repo
Judging for the Anthropic Virtual Hackathon happens in two stages:
-
Date: April 26th - April 27th
-
How it works:
-
Judges review your submitted projects asynchronously via the judging platform.
-
Each team will have uploaded:
-
A short demo video (3 minute maximum)
-
Open Source Project repository / code
-
Written summary (100–200 words)
-
-
Judges independently evaluate projects using standardized criteria.
Judging Criteria:-
Impact (30%) - What's the real-world potential here? Who benefits, and how much does it matter? Could this actually become something people use? Does it fit into one of the problem statements listed above?
-
Demo (25%) — Is this a working, impressive demo? Does it hold up live? Is it genuinely cool to watch?
-
Opus 4.7 Use (25%) - How creatively did this team use Opus 4.7? Did they go beyond a basic integration? Did they surface capabilities that surprised even us?
-
Depth & Execution (20%) — Did the team push past their first idea? Is the engineering sound and thoughtfully refined? Does this feel like something that was wrestled with — real craft, not just a quick hack?
-
-
After evaluation, the team aggregates scores to determine the Top 6 projects for the final round
-
-
Date & Time: April 28th, 12:00PM EST
-
Format:
-
During the live session, pre-recorded demos from each team will be played (3 minutes per team).
-
Judges will deliberate after all demos to determine the 1st, 2nd, and 3rd place along with extra prizes.
-
The top three and special prize winners will be announced during the closing ceremony at 1:45PM EST.
-
-
Submit your project below:
[linkEmbed]
-
Required for submission:
-
3-minute demo video (YouTube, Loom, or similar)
-
GitHub repository or code link
-
Written description / summary
-
-
Deadline: April 26th, 8:00 PM EST
-
Make sure your project was built entirely during the hackathon; no pre-existing work is allowed.
-
1st Place: $50,000 in Claude API Credits
-
2nd Place: $30,000 in Claude API Credits
-
3rd Place: $10,000 in Claude API Credits
-
Most Creative Opus 4.7 Exploration: $5,000 in Claude API Credits
- For the project that treated Opus 4.7 as a creative medium, not just a tool. We're looking for the project with a voice, a point of view — something expressive, playful, strange, or alive. The one that made us feel something.
-
The “Keep Thinking” Prize: $5,000 in Claude API Credits
- For the project that didn't stop at the first idea — and landed somewhere nobody saw coming. We're looking for the team that found a real-world problem nobody thought to point Claude at. The one that changes how we think about where this technology belongs.
-
Best use of Claude Managed Agents: $5,000 in Claude API Credits
- For the team that leveraged the Claude platform the best. We're looking for the project that best uses Managed Agents to hand off meaningful, long-running tasks — not just a demo, but something you'd actually ship.