Everything that decides whether your AgentHack 2026 entry counts and how it is scored — team rules, the 24-hour build window, allowed tools, submission requirements and the four judging criteria. AgentHack 2026 runs 15–16 August 2026, 1 PM to 1 PM PKT.
An AgentHack 2026 team is 1 to 3 builders — no more. You may compete solo. Each person can only be on one team, and the team registered at aiseason.tech/hackathon/apply is the team that is judged. Team members can be from different cities, universities or countries since AgentHack is fully online. One entry fee of PKR 2,250 early-bird covers the whole team.
The 24-hour build window
Building starts at 1:30 PM PKT on 15 August 2026, right after the kickoff and problem reveal, and code must be submitted by 11:00 AM PKT on 16 August 2026. Nothing written before kickoff may be used. There are no pre-set tracks and no pre-built projects — every team starts from an empty repository on the same problem statement at the same moment.
What you are allowed to use
Anything public and available to everyone. Any LLM provider (OpenAI, Anthropic, Google, open models), any agent framework (LangChain, LangGraph, MCP, CrewAI, or none), any language, any cloud, any open-source library, and AI coding assistants like Claude Code, Cursor or Copilot. What you may not use is private code written before kickoff, or work done by someone outside your registered team.
What you submit
By 11:00 AM PKT on 16 August 2026 every AgentHack team submits three things: a link to the source code repository, a demo video showing the agent actually working, and a short written explanation of what you built, how the agent plans and uses tools, and what you would do next. Submissions after the deadline are not scored.
How the 3-minute presentation works
Each AgentHack team gets exactly 3 minutes in front of the judges: 1 minute of live working demo, 1 minute of pitch explaining what you built and why it matters, and 1 minute of Q&A. The clock is hard. Practice the demo — a recorded fallback is allowed if your live run fails, but a live demo scores better on execution.
The four judging criteria
AgentHack 2026 is scored on agentic depth (does the system genuinely plan, call tools and act, rather than being a single prompt in a wrapper), execution (does it work end to end and is it well built for 24 hours), impact and originality (is the idea novel and genuinely useful), and demo clarity (can you show and explain it crisply). Judges score independently and the totals decide the top three.
Plagiarised code, submissions built before kickoff, help from people outside the registered team, and harassment of participants, mentors or judges all mean disqualification. AgentHack is a friendly event — ask mentors for help freely, that is what they are there for. Teams keep full ownership of everything they build.
Frequently asked questions
What are the AgentHack 2026 rules?
AgentHack 2026 teams are 1 to 3 builders, all code must be written inside the 24-hour window starting 1:30 PM PKT on 15 August 2026, any LLM and framework is allowed, and each team submits a code repo, a demo video and a written explanation by 11:00 AM PKT on 16 August 2026.
Can I use AI coding assistants at AgentHack?
Yes. Claude Code, Cursor, GitHub Copilot and any other AI coding assistant are allowed at AgentHack 2026. The event is about what your agent does, not about typing speed. What is banned is code written before kickoff or by anyone outside your registered team.
Can I start building before AgentHack begins?
No. AgentHack 2026 has no pre-set tracks and the problem statement is only revealed at kickoff at 1:00 PM PKT on 15 August 2026. Every team starts from an empty repository. Using code written before kickoff is grounds for disqualification.
What do I have to submit at AgentHack 2026?
Three things by 11:00 AM PKT on 16 August 2026: a link to your source code repository, a demo video showing the agent working, and a short written explanation of what you built and how the agent plans and uses tools.
How long is the AgentHack presentation?
Three minutes per team — a 1-minute live working demo, a 1-minute pitch and a 1-minute Q&A with the judges. Presentations run from 1:00 PM PKT on 16 August 2026, right after the 24-hour build window closes.
What framework should I use for AgentHack?
Whatever fits the problem. LangChain, LangGraph, MCP, CrewAI, or direct API calls with no framework at all are equally acceptable at AgentHack 2026. Judges score agentic depth — whether the system really plans, uses tools and acts — not which library you picked.