Your AI woke up in Vegas with no memory of last night. Build AI that doesn't forget with Cognee's self-hosted, hybrid graph-vector memory layer.
The Jackpot
$10,000 in prizes. The house always remembers, and so will you.
Grand Prizes for All Team Members
Maximum team size of 4
Apple MacBook Neo
One per team member, or the equivalent cash amount, for the best build on the open-source Cognee
Apple iPhone 17
One per team member, or the equivalent cash amount, for the best build on Cognee Cloud
Top winners get job interviews at Cognee
Showcase your skills directly to the team building the memory layer for AI. (Interviews do not guarantee a job, see the rules.)
Open Source Track: $100 per PR · Top 20 submissions
Find issues on the Cognee GitHub repository and contribute to the open-source project. The top 20 PR submissions earn $100 each.
Side Track · Best Blogs
Write about your build, your journey, or how Cognee gives AI a memory. The best blogs win a Keychron mechanical keyboard worth $120.
Side Track · Social Buzz
Share your progress on socials and tag @wemakedevs and Cognee. The top 10 social media posts get exclusive swag shipped to them.
Your AI Has a Hangover
It wakes up every morning with no memory of last night. We're here to fix that.
The Problem
When you call an LLM, every request is stateless. It doesn't remember what happened in the last session, and it quickly spills out of its context window. So your agent forgets the groom, loses the plot, and wakes up on the roof asking “where's my context?”
The Solution
Use Cognee to build AI that does not forget. Give your agents a permanent, self-hosted, hybrid graph-vector memory layer so they can learn, adapt, and carry context across infinite sessions. No more amnesia. No more “what happened last night?”
Powered By Cognee
Cognee is an open-source, self-hosted memory layer for AI agents. It turns text, files, and URLs into a hybrid graph-vector knowledge store, so your agents can remember, recall, improve, and forget, carrying context across infinite sessions. Your mission: build AI that doesn't wake up in Vegas with no memory of last night.
Free Cognee Cloud Developer plan ($35 value)
Sign up to Cognee Cloud for free and unlock the Developer plan with code COGNEE-35 for free access.
How to sign up
Four Operations. Total Recall.
Cognee's memory API is dead simple. If The Wolfpack had this, they'd have found Doug in one line of code.
remember()Ingest text, files, and URLs and permanently structure them into the knowledge graph.
recall()Query memory, Cognee automatically routes the search between semantic similarity and deep graph traversals.
improve()/ memifyRun post-ingestion enrichment, prune stale nodes, and adapt weights based on user feedback.
forget()Surgically prune or delete datasets when they're no longer needed.
import cognee
# 1. Give your agent a memory
await cognee.remember("Doug is the groom. The wedding is Sunday.")
await cognee.remember(file="vegas_receipts.pdf")
# 2. Ask it anything, across infinite sessions
answer = await cognee.recall("Where is Doug?")
# 3. Let the memory get smarter over time
await cognee.improve() # a.k.a. memify
# 4. Surgically forget what no longer matters
await cognee.forget(dataset="last_nights_mistakes")The Theme Is Open. Anything Goes.
Build anything you want: agents, apps, tools, games, automations, as long as you use Cognee for memory. The examples below are just inspiration to get the ideas flowing. They are examples only, not requirements.
Personal Memory Agents
Assistants that remember every conversation, preference, and decision across infinite sessions. Your AI finally stops asking the same question twice.
Research & Knowledge Copilots
Ingest papers, docs, and the web into a living knowledge graph, then recall answers with deep graph traversals. Think Andrej Karpathy Wiki, but yours.
Never-Forget Workflows
Automations and pipelines that carry context between runs. Build agents that learn from yesterday and act smarter today.
Self-Improving Agents
Use improve()/memify to enrich memory and adapt weights from feedback so your agent gets sharper the more it's used.
Support & Customer Memory
Support bots that remember a customer's full history, past tickets, and context, no more 'can you repeat your account number?'
Learning & Tutoring Tools
Tutors that track what a learner already knows, adapt to their pace, and build a personalized knowledge map over time.
Got a wilder idea? Even better. Surprise us, the only rule is that Cognee powers the memory.
Don't Build It Alone
You don't have to build custom agent frameworks from scratch. Wire up your stack, plug into your favorite tools, and ship faster.
Setup & Configuration
Configure your LLM providers, vector stores, graph stores, and everything else you need to run Cognee your way.
Claude Code
Give Claude Code local project memory so your IDE agent never forgets your codebase.
OpenClaw
Drop Cognee memory into OpenClaw with the official npm package or ready-made skills.
Judging Criteria
Potential Impact
How effectively does the project address a meaningful problem or unlock a valuable use case with persistent AI memory?
Creativity & Innovation
How unique is the idea? Does it push the boundaries of what's possible when an agent never forgets?
Technical Excellence
How well is the project implemented? Does it demonstrate strong engineering practices and clean, maintainable code?
Best Use of Cognee
How deeply and effectively does the project lean on Cognee's memory lifecycle APIs and its hybrid graph-vector memory layer?
User Experience
Is the project intuitive to use? Does it provide a polished experience that users would actually want to adopt?
Presentation Quality
How clearly is the project presented? Do the demo, README, and submission communicate the problem, solution, and impact?
Frequently Asked Questions
Still have questions? Join the community and get help in real time.
Join the Discord