Delegate with context
A supervisor–worker architecture built with Google ADK and LangGraph lets specialized agents discover tools, invoke MCP servers, and coordinate multi-step workflows.
PROJECT 01 Multi-agent intelligence
CortexMany agents. One thoughtful answer.
A supervisor–worker system that connects specialized agents, tools, and knowledge to solve multi-step problems.
THE CONTEXT
Complex questions rarely fit neatly into a single tool or data source. Cortex coordinates specialized agents around a shared task, combining orchestration with grounded retrieval.
Connect the knowledge
An agentic Graph-RAG pipeline combines vector search and Neo4j. Query decomposition, re-ranking, and reflection loops support grounded answers with citations.
Make quality observable
Guardrails, automated evaluation, distributed tracing, and inference cost optimization support containerized deployments on GCP.
Coordinated reasoning, grounded in connected knowledge.
Talk through the detailsNEXT EXPLORATION