Difference between ralphex, /simplify and /batch introduced in Claude Code 2.1.63

[paskal]

Just wanted to share, here is claude code itself take on the differences:

Ralphex vs /batch vs /simplify — Claude Code automation compared

All three tools orchestrate Claude Code agents to transform code, but they target very different workflows.

Aspect	/batch	/simplify	ralphex
Purpose	Apply one change across many files	Review + fix recent changes	Execute full plan + review pipeline
Input	One-line instruction	git diff (your recent changes)	Markdown plan file with tasks + checkboxes
Trigger	Proactive — you describe the change	Reactive — run after making changes	Proactive — run before writing any code
Agents	N parallel (one per file)	3 parallel specialised reviewers	5 reviewers (phase 2) + 2 (phase 4)
Review focus	None — executes only	Reuse · Quality · Efficiency	Bugs · Goals · Tests · Simplification · Docs
Git integration	No	No	Full: branch, commits, rebase, squash
Fixes	The requested change	Issues found in your diff	Issues found in its own output
Iteration	Single pass	Single pass	Loops until clean
External tools	None	None	OpenAI Codex, Telegram/Slack/email notifications
Autonomy	One-shot	One-shot	Walk-away

Similarities

• All three orchestrate multiple Claude Code agents working in parallel
• All target code transformation/improvement across a codebase
• All reduce manual supervision for repetitive tasks
• Fresh agent sessions prevent context window degradation

Mental model

• /batch — "do this thing everywhere"
• /simplify — "review what I just did and clean it up"
• ralphex — "build this entire feature, review it yourself, then hand me a clean branch"

/simplify is essentially the manual, on-demand version of ralphex's automated review phases. /batch has no equivalent in ralphex — it solves a different problem (uniform changes across many files rather than sequential task execution).

[umputun]

/simplify is essentially the manual, on-demand version of ralphex's automated review phases.

I have not tried /batch, but /simplify is most definitely not the same as the review phase for several reasons:

1. /simplify is a single phase (which is correctly noted in the table), versus an N-phase loop with a second opinion.
2. Three included agents are analyzing something, but they seem to be extremely forgiving. I have tried it on the same branch I used ralphex --review after, and it reported "all good." The Ralphex review did multiple loops, and each one found subtle edge cases worth fixing, as well as several styling violations. Note: I'm using custom review prompts, invoking my own specialized agent.