Give your agents an understanding of your software.

Give your agent a feature, behavior, or issue to explore. Driver identifies the relevant codebases so it can start in the right place, even when you don’t know which repository owns the work.

Your agent can use connected codebases you have access to, even when they aren’t cloned locally.

Give your agent an overview of the system’s components, their responsibilities, and how they connect. Help it place a feature or behavior in the wider architecture.

Your agent can move from that overview to focused explanations and code references for the relevant component.

Help your agent locate the implementation behind a behavior. Descriptions of folders and files give it a starting point without reading every file to understand its purpose.

Your agent can follow a behavior across codebases, with each lookup scoped to the branch you need.

Give your agent the surrounding code needed to understand an implementation: dependencies, mirrored definitions, registration patterns, and files that change together.

Help it follow those relationships beyond the file it started with.

Help your agent assess a proposed change by tracing indirect dependencies and identifying candidate tests. Give it evidence to decide what else needs review and verification.

Driver automatically retains findings, corrections, and resolutions from investigations. Your agent can retrieve that knowledge when a related question comes up.

Keep prior findings connected to their supporting evidence, so your agent can build on what is known and see what still needs verification.