AI agents

Hand a scan to Claude Code, Codex, or Cursor without wiring anything up.

Rust Doctor exists for codebases where an agent writes most of the Rust, so the report is built to be handed to one. There is no server to run and nothing to configure.

#The handoff

From the interactive report, pick the handoff entry. Rust Doctor composes a bounded brief of the top findings, with rule identifiers, file locations and the fix guidance the catalog carries, and copies it to your clipboard for Claude Code, Codex, or Cursor.

The brief is capped at 12 KB, covers the three highest-value groups and up to 24 locations, and includes the exact command to rescan afterwards. That cap is the point: an agent that receives every finding fixes none of them well.

#The JSON path

For an agent that reads files rather than prompts:

rust-doctor --json --yes . > report.json

--yes is what tells Rust Doctor its reader is not a terminal. The report is workspace-relative and carries no environment data, so it can go straight into a context window or an issue. See JSON report.

#Scoping the work

Ask for the change and nothing else:

rust-doctor --scope baseline --base origin/main --json --yes .

The agent then sees only what its own change introduced, which is usually the question you meant to ask.

#A caveat worth stating

A scan runs cargo clippy inside the workspace, and Cargo executes that workspace's build scripts and procedural macros. Point Rust Doctor at code you trust, whoever wrote it: an agent that pulled in a dependency you have not read is exactly the case this warning is about.