CockroachDB × AWS Hackathon
Agents that think together.
Agents that remember together.
When one agent learns, they all learn. HiveMind gives AI agents a shared, persistent memory — so knowledge compounds across agents, across runs, across time.
Coderwrites
Shared Memorypersists
Reviewerreads
Deployerdeploys
Powered by CockroachDB distributed vector indexing · Always-on · Zero data loss
How it works
Three steps. Persistent memory. Emergent intelligence.
Agents write memories
Every finding, critique, fix, and deployment plan is written to shared memory in CockroachDB — with vector embeddings for semantic search.
Agents read memories
Before acting, each agent searches shared memory for relevant context — past code, review patterns, deployment history, and injected issues.
The hive compounds knowledge
Memory persists across runs, restarts, and failures. The hive gets smarter every day. What one agent learns, all agents benefit from.
The memory difference
Without shared persistent memory, agents are just stateless chatbots.
Without shared memory
- Agents are isolated — no shared knowledge
- Memory is ephemeral — lost on restart
- Each agent repeats the same mistakes
- No cross-agent learning or collaboration
- Context window is the only memory
With HiveMind
- Shared persistent memory across all agents
- Memory survives restarts, failures, and regions
- Agents learn from each other's past experiences
- Semantic search finds relevant knowledge instantly
- Knowledge compounds — the hive gets smarter daily
Built on
CockroachDBDistributed SQL with vector indexing
Amazon Web ServicesS3, Bedrock, cloud infrastructure
OpenAI GPT-5.6Agent reasoning and code generation
MCP ServerDirect agent ↔ database connection