Agents
Agents combine models, instructions, tools, state, and an execution loop so the system can decide how to pursue an outcome. The important design choice is not whether to use the label agent, but how much control the model receives and where deterministic boundaries remain.
Questions to answer
Resolve these before adding tools, frameworks, or automation.
- What decisions may the model make during a run?
- Which outcomes require deterministic checks or human approval?
- How will the run stop, recover, and report partial progress?
Implementation lifecycle
Build the evidence in this order
- 01
Define
Write the outcome, allowed decisions, stop conditions, and accountable owner.
- 02
Build
Use the simplest loop and smallest tool surface that can complete a representative task.
- 03
Verify
Trace the run and test success, refusal, timeout, and partial-failure paths.
Artifacts to maintain
- agent definition
- run policy
- tool registry
- completion contract
Continue through AgentAtlas
Evidence ledger
Primary documentation behind this guide
Anthropic
Building effective agents
The distinction between fixed workflows and model-directed agents, plus the recommendation to add complexity only when measured outcomes justify it.
OpenAI
Running agents
Agent run loops, state management, continuation strategies, and the operational boundary between the SDK and application-owned state.
Next topic
Skills