Layer 01/08
Create Agents
Compose production-ready agents from models, tools, memory, and knowledge — without stitching together a stack.
The problem
Building an agent usually means gluing a model, a vector store, a tool layer, and orchestration code together, then re-solving the same plumbing for every new use case. It is slow, brittle, and impossible to govern consistently.
MODEL
GluedVECTOR STORE
GluedTOOL LAYER
GluedORCHESTRATOR CODE
Glued
What it does
Assemble agents from reusable building blocks: a model (LLM or private SLM), tools and function calls, memory, and grounded knowledge.
Configure behaviour with system prompts, role definitions, and policies — versioned like code.
Attach knowledge bases and RAG/GraphRAG sources so agents answer from your data, not their training set.
Reuse components across agents so a fix in one place propagates everywhere.
How it connects
Agents created here flow into Design Workflows for orchestration, Test Agents for validation, and inherit the controls defined in Access & Governance and the AI Firewall.