About the Challenge
Ready to build AI agents that don’t forget? 🔥
MemVerge is hosting the Memories That Last – AI Agents Hackathon this October, where you’ll explore the future of stateful AI and compete for $4,000 in cash prizes.
At the heart of this challenge is MemMachine, our open-source AI memory system that transforms ordinary chatbots into agents that learn, adapt, and remember. Instead of starting from scratch every time, your AI will grow smarter with every interaction.
Whether you’re a coder, founder, researcher, or student, this hackathon is your launchpad to:
- 💡 Turn bold ideas into working MVPs in just 48 hours
- 🤝 Team up with innovators and builders who share your vision
- ⚡ Get hands-on support from the MemVerge team before, during, and after the event
If you’ve ever wanted to build AI that feels more human — agents with memory that lasts — this is your chance. Bring your creativity, ship your solution, and make your mark.
Read more about MemMachine here.
Requirements
What to Build
Form teams to build a chat-based assistant or AI agent in 48 hours or less that utilizes MemMachine as the AI memory and context layer.
What to Submit
Each team will submit their final project to GitHub at the end of the hackathon along with a 3 minute or less demo video. Projects will be judged and winners will be announced on Friday, October 10th, 2025.
Prizes
1st Place
2nd Place
3rd Place
Devpost Achievements
Submitting to this hackathon could earn you:
Judges
Jing Xie
Vice President AI Memory
Gauri Nagavkar
Senior Developer Advocate, Bioinformatics and AI
Aulene De
Senior Software Engineer, Meta
Judging Criteria
-
Usage & Memory Quality
How well does the project integrate MemMachine as a persistent AI memory layer? Does the agent recall relevant information at the right time, adapt across sessions, and clearly demonstrate memory in action? -
Functionality & Reliability
Is the solution stable, reproducible, and technically sound? Judges will check if persistence works after restarts, and whether the app is easy to run (e.g., via Docker or scripts). -
User Experience & Transparency
Is the agent intuitive to use, and does it clearly show how memory influences behavior? Bonus points for features like memory inspection, editing, or forgetting. -
Innovation & Presentation
Does the project apply persistent memory to a meaningful problem in a creative way? Is the submission well-documented with a clear README and demo video that proves memory is working (teach → restart → recall)?
Questions? Email the hackathon manager
Tell your friends
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
