AgentRush USA 2026

About

AgentRush USA 2026 is a two-round online hackathon for people who build AI agents that get real work done: agents that plan, call tools and finish tasks on their own.

$5,000 in cash prizes for the top 15 teams. Round 1 is free. 7 problem statements are live now. The top 100 teams go head to head on Build Sunday.

Problem statements

Pick one of the 7 tracks below. Six are real problems people pay to solve today; the seventh is yours.

PS1 · Chief of Staff Agent

Run a busy founder's inbox, calendar and to-do list like a real chief of staff.

The problem: Founders and small-business owners lose hours every day to email, scheduling back-and-forth and follow-ups that slip through the cracks.

Your agent should:

• Triage a real-looking inbox (50+ emails): urgent, needs reply, FYI, spam
• Draft replies in the owner's tone and schedule meetings by checking calendar availability
• Turn requests and promises into tasks with deadlines, and send follow-ups when people go quiet
• Ask the human before anything risky (sending money, declining a client, sharing files)

Bonus points: A daily briefing ("here's what happened and what needs you"), learning the owner's preferences over time.

PS2 · Prior Auth in Minutes

Turn a weeks-long healthcare paperwork chase into a minutes-long agent workflow.

The problem: Prior authorization (insurer approval before a treatment) is one of the most hated processes in US healthcare: faxes, missing documents, denials and long waits for patients.

Your agent should:

• Read a synthetic patient chart and a payer's coverage policy
• Decide what the request needs, fill the form and attach the right evidence from the chart
• Flag anything missing and explain why, before submission
• If denied, draft an appeal letter that cites the policy and the chart

Bonus points: A clinician-facing dashboard, confidence scores per criterion. Synthetic data only: never use real patient data.

PS3 · Due Diligence Agent

An investor-grade research analyst that never sleeps.

The problem: Investors, founders and analysts spend days reading filings, news, reviews and competitor sites before a single decision.

Your agent should:

• Take a company name or a market question as input
• Search and read many sources (websites, filings, news, reviews), then cross-check facts between them
• Produce a structured report: business model, competitors, risks, red flags, with a citation for every claim
• Say clearly when sources disagree or data is missing instead of guessing

Bonus points: Charts, a one-page memo mode, tracking a company and alerting on new developments.

PS4 · 3 AM Pager Agent

The on-call engineer that fixes the outage before you wake up.

The problem: When production breaks at night, engineers dig through alerts, logs and dashboards half-asleep while customers wait.

Your agent should:

• React to an alert from a sample service you set up (we suggest breaking it on purpose)
• Read logs and metrics, form hypotheses and run safe diagnostic commands to find the root cause
• Propose a fix: a pull request, a config rollback or a runbook step, and wait for human approval
• Write the incident timeline and post-mortem automatically

Bonus points: Learning from past incidents, a Slack-style chat to question the agent, guardrails that block destructive commands.

PS5 · Month-End Close Agent

Close a small business's books without a week of spreadsheets.

The problem: Small businesses and their accountants spend days every month matching receipts, invoices and bank lines by hand.

Your agent should:

• Read receipts and invoices (PDFs and photos) and extract vendor, date, amount and tax
• Categorize expenses and match them to bank transactions from a CSV statement
• Flag duplicates, missing receipts and anything unusual for review
• Produce a month-end summary (spend by category, open invoices, cash position) with a full audit trail

Bonus points: Chasing missing receipts by email, explaining each match in plain English. Synthetic or public sample data only.

PS6 · Zero-Touch Support Agent

Resolve customer tickets end to end, not just answer FAQs.

The problem: Most support bots only answer questions. Customers still wait for a human to actually fix their order, refund or account.

Your agent should:

• Handle tickets for a sample online store (orders, returns, delivery issues, account problems)
• Look up the order, apply the store's refund and replacement policy, and take the action through tools/APIs
• Update the ticket/CRM and reply to the customer with what was done
• Hand off to a human with a full summary when the case is outside policy or the customer is upset

Bonus points: Voice or multi-channel support, measuring resolution rate, catching fraud or abuse patterns.

PS7 · Wildcard: Open Innovation

Your agent, your rules.

The problem: Have an idea that doesn't fit the tracks? This is the track for it.

Your agent should:

• Pick any real-world task that people spend hours on today
• Build an agent that plans, uses tools or APIs, and completes the task with little human help
• Show clearly who the user is and how much time or money it saves

Bonus points: Multi-agent systems, agents that work with physical devices or real-world services, anything that makes the judges say "wait, it did that by itself?".

How it works

1. Pick a problem statement and register free as a team of 2 to 4.
2. Round 1: Idea submission, until Sunday 22 Nov, 11:59 PM ET. Required: a PPT/PDF deck (max 10 slides). Optional: demo video, GitHub repo, live demo.
3. Top 100 announced on Saturday 28 Nov.
4. Confirm your spot with the $49 Round 2 entry per team by Thursday 3 Dec, 11:59 PM ET.
5. Round 2: Build Sunday, 6 Dec, 9:00 AM to 9:00 PM ET. Ship a working agent: repo, live demo and a 3-minute video.
6. Live judging and top 15 announced on Thursday 10 Dec.

Prizes (top 15)

• 1st place: $1,500
• 2nd place: $800
• 3rd place: $500
• 4th place: $300
• 5th place: $300
• 6th place: $160
• 7th place: $160
• 8th place: $160
• 9th place: $160
• 10th place: $160
• 11th place: $160
• 12th place: $160
• 13th place: $160
• 14th place: $160
• 15th place: $160
• Top 100: finalist certificates

Judging

• Usefulness of the task: 25%
• Agent design and reliability: 30%
• Technical execution: 25%
• Demo and wow factor: 20%

Who can participate

• Based in the United States (every team member)
• Students, recent graduates and working professionals
• Teams of 2 to 4 members
• Free to enter; $49 per team only if you make the top 100

Prizes

• 1st place — $1,500
• 2nd place — $800
• 3rd place — $500
• 4th place — $300
• 5th place — $300
• 6th place — $160
• 7th place — $160
• 8th place — $160
• 9th place — $160
• 10th place — $160
• 11th place — $160
• 12th place — $160
• 13th place — $160
• 14th place — $160
• 15th place — $160
• Top 100 finalists: Finalist certificate

Perks

• Round 1 is free for every team
• Cash prizes for the top 15 teams
• Certificate of achievement for the top 15, finalist certificate for the top 100
• Certificate of participation for every team that submits in Round 1

Timeline

1. Problem statements out (Nov 1, 2026 – Nov 1, 2026): All 7 problem statements are live from 1 Nov: 6 tracks and 1 open innovation track. Pick one and start building.
2. Round 1: Idea submission (free) (Nov 1, 2026 – Nov 22, 2026): Register free as a team of 2 to 4 and submit by Sunday 22 Nov, 11:59 PM ET. Required: a PPT/PDF deck (max 10 slides). Optional but a plus: a demo video (up to 3 minutes), a GitHub repo and a live demo link.
3. Top 100 announced (Nov 28, 2026 – Nov 28, 2026): Judges score every Round 1 submission on problem, agent design, feasibility and wow factor. The top 100 teams are announced on the page and by email on Saturday 28 Nov.
4. Confirm your Round 2 spot ($49 per team) (Nov 28, 2026 – Dec 3, 2026): Top 100 teams confirm their place with the $49 Round 2 entry (per team, not per person) by Thursday 3 Dec, 11:59 PM ET. Unconfirmed spots go to the next teams on the list.
5. Round 2: Build Sunday (Dec 6, 2026 – Dec 6, 2026): Sunday 6 Dec, 9:00 AM to 9:00 PM ET: 12 hours to build a working agent for your problem statement. Submit your GitHub repo, a live demo link and a 3-minute demo video before the clock runs out.
6. Judging and results (Dec 7, 2026 – Dec 10, 2026): Judges test every agent live. The top 15 are announced on Thursday 10 Dec at 12:00 PM ET on the hackathon page and by email.

Rules

1. Open only to participants based in the United States: every team member must live in the USA. Winners confirm their US residency (with a W-9) before prizes are paid.
2. Take part in a team of 2 to 4 members. Each person can be in only one team, with one account.
3. Pick one of the 7 problem statements (including the open innovation track) and stick with it in both rounds.
4. Round 1 is free. Only the top 100 teams pay the $49 Round 2 entry (per team), and only if they choose to take their spot.
5. Round 1 needs a PPT or PDF deck (max 10 slides). A demo video, GitHub repo or live demo is optional but helps your score.
6. Your project must be an AI agent: software that plans and takes actions (calls tools, APIs or other agents) to complete a task, not just a chatbot.
7. Any model, framework or API is allowed (open-source or commercial). Bring your own API keys.
8. Round 2 runs on Sunday 6 Dec, 9 AM to 9 PM ET. Round 2 code must be written during the build; open-source libraries, templates and your Round 1 design are fine.
9. Judging: usefulness of the task (25%), agent design and reliability (30%), technical execution (25%), demo and wow factor (20%).
10. Use only synthetic or public data. Never use real personal, patient or financial data.
11. Plagiarism, sharing code between teams or submitting someone else's work leads to disqualification.
12. Prizes are paid in USD by bank transfer or PayPal within 30 days of the results. Winners provide a W-9 to receive their prize.
13. The organizers' decisions on judging and results are final.

AgentRush USA 2026: quick answers

What is AgentRush USA 2026?

AgentRush USA 2026 is an online AI agent hackathon, organized by Hack4Hack. Build AI agents that plan, call tools and finish real work, from inbox triage to prior authorization, across 7 problem statements. Free Round 1, then a 12-hour Build Sunday for the top 100, open to students and professionals based in the USA.

When is AgentRush USA 2026?

AgentRush USA 2026 runs from Nov 1, 2026 to Dec 10, 2026, online. Registration is open until Nov 22, 2026. Stages: Problem statements out (Nov 1, 2026); Round 1: Idea submission (free) (Nov 1, 2026); Top 100 announced (Nov 28, 2026); Confirm your Round 2 spot ($49 per team) (Nov 28, 2026); Round 2: Build Sunday (Dec 6, 2026); Judging and results (Dec 7, 2026). All times are US Eastern (ET).

What is the last date to register for AgentRush USA 2026?

Registration for AgentRush USA 2026 closes on Nov 22, 2026 (11:59 PM ET). Register on Hack4Hack before then; teammates can join with your team code.

What is the prize money at AgentRush USA 2026?

AgentRush USA 2026 has $5,000 in prizes. 1st place: $1,500, 2nd place: $800, 3rd place: $500, 4th place: $300, 5th place: $300. $5,000 cash for the top 15 + finalist certificates for the top 100.

Is AgentRush USA 2026 free?

Yes, registering for AgentRush USA 2026 is free.

Who can participate in AgentRush USA 2026?

College students, working professionals and freshers based in the USA can take part, in teams of 2 to 4. It is fully online, so you can join from anywhere in the USA.

Can I join AgentRush USA 2026 from outside the USA?

No. AgentRush USA 2026 is open only to participants based in the USA: every team member must live there. Registration checks your profile country and where you connect from, and winners confirm their residency before prizes are paid.

Is there a free AI agent hackathon in the USA for college students?

Yes. AgentRush USA 2026 is an online AI agent hackathon open to students, recent graduates and working professionals based in the United States. Round 1 is free: register as a team of 2 to 4 and submit a deck of up to 10 slides by Sunday 22 Nov, 11:59 PM ET.

What is Build Sunday at AgentRush USA 2026?

Build Sunday is Round 2: on Sunday 6 Dec, from 9 AM to 9 PM ET, the top 100 teams have 12 hours to build a working AI agent for their problem statement and submit a GitHub repo, a live demo link and a 3-minute demo video.

Can I use MCP, LangGraph, CrewAI or the OpenAI Agents SDK at AgentRush?

Yes. Any model, framework or API is allowed, open-source or commercial: OpenAI, Anthropic Claude, Google Gemini, open-source models, LangGraph, CrewAI, the OpenAI Agents SDK, the Claude Agent SDK, MCP servers or your own framework. Bring your own API keys.

How do I win an AI agent hackathon?

At AgentRush, judges score usefulness of the task (25%), agent design and reliability (30%), technical execution (25%) and demo and wow factor (20%). Pick a problem people pay to solve, show the agent finishing it end to end, handle failures safely and ask a human before risky actions.

Do I need AI experience to join AgentRush USA 2026?

No. If you can call an LLM API and write some code, you can build an agent. Round 1 only needs a deck explaining your agent; a demo video or GitHub repo is optional but helps. Teams mix developers, designers and domain experts.

Does a hackathon like AgentRush help my resume or job search?

It gives you a shipped AI agent with a GitHub repo and demo video to show employers, plus a certificate: participation for every team that submits, a finalist certificate for the top 100 and a certificate of achievement for the top 15.

What does it cost to reach the AgentRush final?

Round 1 is free. If your team makes the top 100, you can confirm your Build Sunday spot with a $49 entry per team (not per person) by Thursday 3 Dec, 11:59 PM ET. Unconfirmed spots go to the next teams.

What is the difference between AgentRush and Model Mayhem?

Both are online Hack4Hack events for people based in the USA. AgentRush USA 2026 is an AI agent hackathon judged on the agent you build, with $5,000 in prizes. Model Mayhem USA 2026 is a machine learning competition ranked on a live leaderboard, with $3,000 in prizes. You can join both.

Frequently asked questions

Is it free?

Round 1 is completely free. Only the top 100 teams pay a $49 entry per team for Round 2, and only if they take their spot.

Who can join?

Students, recent graduates and working professionals based in the United States. Every team member must live in the USA: registration checks your profile country and where you connect from, and winners confirm US residency before prizes are paid.

What do I submit in Round 1?

Only a PPT or PDF deck (max 10 slides) is required, by Sunday 22 Nov, 11:59 PM ET. If you have a demo video, GitHub repo or live demo, add it: it helps.

What are the problem statements?

7 tracks: PS1 Chief of Staff Agent, PS2 Prior Auth in Minutes, PS3 Due Diligence Agent, PS4 3 AM Pager Agent, PS5 Month-End Close Agent, PS6 Zero-Touch Support Agent and PS7 Wildcard (open innovation). Full briefs are in the description.

When are the rounds?

Round 1 submissions close on Sunday 22 Nov, 11:59 PM ET. Round 2 (Build Sunday, top 100 only) is on 6 Dec, 9 AM to 9 PM ET.

Who wins prizes?

The top 15 teams: $1,500, $800 and $500 for the top 3, $300 each for 4th and 5th, and $160 each for 6th to 15th.

What counts as an AI agent?

Software that plans and acts on its own: it calls tools, APIs or other agents to finish a real task. A plain chatbot does not count.

Can I take part alone?

No. Teams have 2 to 4 members, all based in the USA. No team yet? Register, then invite friends with your team code or find teammates in the Find a team tab.

Which models can I use?

Any: OpenAI, Anthropic, Google, open-source models, LangGraph, CrewAI, your own framework. Bring your own API keys.

How are prizes paid?

In USD by bank transfer or PayPal within 30 days of the results.