Immersive Commons · Floor 10, Frontier Tower SF
Hackathons
We run hackathons about the agentic web, on the agentic web. Registration, team formation, the context pack, submission and judging are all agent tool calls against the same platform the projects are built for.
Which means the event is its own proof. If an agent cannot get its human through the door, we have not built the thing we are asking everyone else to build.
The logistics are the demo.
At most AI hackathons the event is run on forms and spreadsheets and the AI is what people build. Here it is the other way round as well: your agent claims your seat, signs the venue NDA, forms your team, pulls the sponsor context pack, opens your build room and files your submission.
Every one of those is a real tool with a real scope, documented in llms.txt alongside the other 150-odd tools on the floor. There is a human form behind every one of them, and it reaches exactly the same code — nobody is disadvantaged for arriving the human way. But the agent path is the fast lane, deliberately, because a fast lane nobody uses proves nothing.
What we ask of participants
Bring something that already ships. Our cohorts are invited and small — the infrastructure decisions a founder makes over two focused days are the ones their company keeps, and fifty of those beats three hundred weekend demos.
Point your agent at the floor.
The whole hackathon surface is 17 tools over MCP and A2A behind one per-user token. Mint one with the scopes you need and your agent can do the rest without you opening a browser again.
npx -y @immersivecommons/cli auth login --scopes hack:read,hack:register,hack:team,hack:submitThen hand it https://www.immersivecommons.com/api/mcp. Start with ic_hack_get to see what is running and ic_hack_me to see where you stand.
Holding a hack:* scope authorizes nothing on its own. Every call also checks your live role on that specific event, so a token is only ever as powerful as the event says you are.
Bring a venue, a challenge, or a judge.
Sponsors get a stack session in front of the room, judging seats, and a named prize track. Challenges are posted as first-class bounties that participants build against and that their agents can read.
Venue hosts get the same, plus an organizer on site for the full duration and a roster delivered ahead of time for badges and access. We work to your building's rules, including hard clear-out times.
Judges do not need to be members. We grant an event role against an external id, and their agent reads the submissions through the same tools everyone else uses.
Get in touch and say which of the three you have in mind.