Connect REA to Codex or Cursor.

Select your client explicitly

REA’s setup supports both Codex and Cursor. Choose the command for the agent you want to connect, review its planned edits, then restart that agent.

# Codex
npx rea-agents@latest setup --client codex

# Cursor
npx rea-agents@latest setup --client cursor

To configure both in one reviewed plan, repeat the client flag. Setup includes matching workflow instructions unless you explicitly disable them.

npx rea-agents@latest setup --client codex --client cursor

Diagnose only the integration you need

npx rea-agents@latest doctor --client codex --json
npx rea-agents@latest doctor --client cursor --json

Check the reported configuration path, package version, and registration state. If the local entry exists but the agent shows no REA tools, restart the active client and inspect its MCP connection status. A doctor result checks configuration; tool discovery happens in the running client.

Start with a bounded prompt

Inspect /path/to/my/application with REA.
First identify the artifact type and available analysis path.
Then trace one feature: importing a JSON settings file.
Explain the parser, validation, and state update.
Separate confirmed evidence from inference.

This prompt is a starting template. Keeping the artifact type and evidence requirements explicit helps the agent select the appropriate tools before it attempts native analysis or runtime capture.

Keep investigations reproducible

Record the REA version, target version, provider, and important source references in your notes. When comparing two releases, use the same question and equivalent inputs. A change in provider output can otherwise look like a change in application behavior.

Before sharing an investigation, review its excerpts for private paths or proprietary data. Local analysis and your agent provider’s data handling are separate concerns.