# Reasoning and cognitive cycles

Use retrieved context and explicit rules to support a task.

Source: https://docs.minds.sh/docs/concepts/reasoning



Reasoning connects available information to a question or action. In Minds, the relevant capabilities include retrieval, symbolic rules, learned signals, and cognitive-cycle controls. They do not all represent a language-model answer endpoint.

## Learned signals and explicit rules [#learned-signals-and-explicit-rules]

Learned representations can find patterns that are difficult to express manually. Explicit rules can make a relationship or constraint inspectable. Combining them is often called **neurosymbolic** processing.

For example, similarity can find candidate records while a rule restricts the result to an authorized project or a valid time range. A useful system preserves the reason a candidate was accepted or excluded.

## A cognitive cycle [#a-cognitive-cycle]

A cycle coordinates processing around current context. Inspect the available configuration, the state of the cycle, and the outcome exposed by its API. Do not assume that enabling a cycle grants an agent authority to run every tool or changes every stored memory automatically.

For an application integration, start with a clear input and one observable result. Add additional steps only after the first path is verified. This keeps reasoning behavior separate from transport, authentication, and persistence problems.

[Reasoning API](/docs/api/instance/reasoning) · [Cognitive cycle API](/docs/api/instance/cognitive-cycle) · [Working memory](/docs/api/instance/working-memory)