Yes. AI agents can follow a project's coding conventions when those conventions are visible in the repository, examples, configuration, or persistent rule files. Explicit rules are more reliable than expecting a model to infer every preference from existing code.
Document naming, directory structure, architecture, error handling, test expectations, dependency policy, and required commands. Put project-specific guidance in AGENTS.md or the tool's supported equivalent and keep it under version control. Use linters, formatters, type checks, and architecture tests to enforce rules that can be automated.
Verdent's VS Code rule-system documentation describes project rules through AGENTS.md, global preferences through VERDENT.md, and plan customization through plan rules. Its rule precedence gives project-specific standards higher priority than global preferences. New VS Code conversations can reference the same durable instructions; confirm current rule inheritance separately when using Manager workers.
Rules still need maintenance. Remove obsolete guidance, resolve contradictions, and include examples for patterns that are difficult to describe. Review the first few generated changes and update the rules when the same error repeats. The goal is not for the agent to imitate accidental legacy code; it is to follow intentional standards that the team can explain and test.
Last verified: July 14, 2026. Pricing, model availability, promotions, and product policies can change; check the linked official source before purchasing or deploying.
