Comparison
WorkflowvsAgent
Workflow
a merge triggers a sequence of stages on someone else's machine, and the sequence is now the only way anything reaches production.
The automated sequence that turns a commit into a tested, published artefact and then into a deployment. It is infrastructure in its own right — it has uptime, capacity, credentials and a blast radius — and it is usually the most privileged system a team operates while receiving the least operational attention. Its slowness is a direct tax on every change, which is why build time is a reliability metric rather than a convenience one.
Full entry →Agent
you stop scripting the steps and let the model choose which tool to call next, in a loop, until it decides it is finished.
A system where a model decides its own next steps in a loop, calling tools, reading results and continuing until it judges the goal met. The defining property is that the control flow is chosen by the model at runtime, not fixed in code.
Full entry →The difference
A workflow chains model calls along a path your code fixes in advance; an agent lets the model choose the path. Workflows are more predictable, debuggable and cheap, and are the right answer more often than the discourse suggests. Reach for an agent only when the path genuinely cannot be predetermined.
Entry: Workflow vs agent →