CockroachDB × AWS Hackathon - Build with Agentic Memory

Overview

CockroachDB and AWS invite developers, engineers, and AI builders to create the next generation of agentic applications. The event focuses on building AI agents with production-grade, persistent memory using CockroachDB as the memory layer, deployed on AWS. The challenge centers on agent memory that survives across regions and failures, enabling agents to remember conversations, contexts, task states, embeddings, and transactional data at scale. The event is online, with activities spanning several weeks from late June through mid August, and features a mix of technical requirements, submission guidelines, and prizes.

Why Agentic Memory

AI agents are increasingly moving from experiments to real production workflows such as coding, data pipelines, incident diagnosis, and high traffic tasks. A core problem is persistent memory that does not fail when an agent restarts. Traditional databases are optimized for human-scale reads and writes, but agentic systems spawn continuously, write constantly, and demand memory that persists across regions and failures without data loss. CockroachDB is positioned as the memory layer for these agents, combining global distribution, durability, PostgreSQL compatibility, and native integration with the agent toolchain through MCP, cloud services, and an open source skills ecosystem. This hackathon invites participants to build on that foundation and demonstrate real world impact.

Theme and Scope

The challenge asks for agentic applications that use CockroachDB as the persistent memory layer, deployed on AWS. Submissions should show how memory is used for state, embeddings, context, and transactional data, illustrating a production ready approach rather than a toy demonstration. Teams should showcase integrations with CockroachDB tools such as distributed vector indexing, MCP Server, and cloud CLI, as well as the Agent Skills repository and LangChain integrations. Participants are encouraged to incorporate AWS services like Bedrock, Lambda, ECS/EKS, S3, SageMaker, or Bedrock agents to power the agent environment.

Format and Tracks

Entries may include public open source repositories with clear README, dependencies, setup, and run instructions. A functional demo app or video (less than 3 minutes) is required to illustrate how the CockroachDB memory layer operates within the agentic system. Submissions should clearly identify which CockroachDB tools and AWS services were used and describe how they interact to enable persistent memory and reliable agent behavior. The judging criteria emphasize Agentic Memory Design, Technical Implementation, Real-World Impact, Production Readiness, and Creativity.

Agenda and Deliverables

• Build an agentic application on CockroachDB memory layer, deployed on AWS
• Use at least two CockroachDB tools (eg MCP Server, Distributed Vector Indexing, ccloud CLI, Agent Skills repo)
• Include a public open source repository with license (MIT or Apache 2.0 recommended)
• Provide a public demo app URL and a short video demonstration
• Provide architectural diagrams (optional) and feedback on CockroachDB AI tools (optional)
• Submit details of AWS services used and how memory is leveraged across sessions and tasks

Prizes and Participation

The challenge offers cash prizes and recognition for top submissions, highlighting the practical impact on AI workflows and agent memory reliability. Participants should be at least the legal age in their country and can form teams of up to five members. The event is managed by Devpost with support from Cockroach Labs and AWS, emphasizing open collaboration and a path toward production ready agentic solutions.

Who Should Attend

Developers, AI engineers, ML and data scientists, and platform builders who want to explore the convergence of distributed databases, agentic memory, and cloud infrastructure. If you want to push the boundaries of what reliable agent memory can accomplish at scale, this hackathon provides a platform to demonstrate production-ready ideas and contribute to the evolving AI toolchain.