Use Code Intelligence reports
Explore code structure, findings, and technical recommendations.
Use Code Intelligence for a focused analysis of your codebase or a technical decision. It produces reports and recommendations you can review before starting implementation.
Ask an answerable question
Open the Code Intelligence workspace and name the repositories and scope. Prefer a concrete question over a request to “analyze everything.”
Map the search request path from the web composer to the backend query.
Identify duplicated filtering logic and recommend a behavior-preserving cleanup.
Include file references and the checks needed to verify the change.Supply constraints that affect the recommendation, such as preserving an API, avoiding a migration, or limiting changes to one package.
Read the report
Review the overview, findings, evidence, and recommended work. Follow source references to confirm the report describes the code version you care about. Ask the agent to distinguish observed structure from inferred runtime behavior.
A static code path can suggest a performance issue without measuring its impact. For performance or reliability decisions, ask what additional traces or experiments would establish the claim.
Turn a recommendation into work
Choose a bounded item, record acceptance checks, and link the implementation task or session. Keep unrelated recommendations as separate work so the resulting diff remains reviewable.
Code Intelligence reports are separate from repository wiki indexing. The repository wiki is currently paused and should not be used as the prerequisite for this workflow.
Continue with Tasks, session creation, or feature availability.