Skip to content

Patterns

RAG assistant

Area 1
Retrieval and grounding
Area 2
Data pipelines, APIs, security
Area 3
Agent implements and tests pipeline
Area 4
Define whose decision the answer supports

Coding agent

Area 1
Model, tools and harness
Area 2
Repository quality and tests
Area 3
Instructions, permissions, orchestration
Area 4
Select appropriate tasks and autonomy

Evals

Area 1
Behavioural datasets and graders
Area 2
CI, regression and observability
Area 3
Verify agent outputs and process
Area 4
Encode desired outcome and risk tolerance

MCP tool

Area 1
Model-accessible capability
Area 2
Contract and security design
Area 3
Agent configuration and permissions
Area 4
Justify access and human authority

Production release

Area 1
Model monitoring and fallbacks
Area 2
CI/CD, reliability and rollback
Area 3
Agent-assisted delivery with review
Area 4
Evidence, risk acceptance and outcome metrics