Help your Claude out.

Save iterations and subscription limits. Get automated, independent review of every Claude Code plan.

Claude Code is great at writing plans. But a single model has blind spots. 1K4's MCP reviewer audits Claude Code plans with up to three different models, cascading findings from each stage and returning them to Claude Code before it exits plan mode. Use it for higher-stakes projects. Fully control the depth and number of iterations. Run once or ad nauseam until consensus. Set budget caps per review. Only use it when you want it.

First plan review on us.

Start setup See it in Action

Belongs in your terminal

Fits right into your plan mode workflow.

● I'll review the plan before we implement.
● 1k4-review – 1k4_review_plan (MCP)(plan_file: "PLAN_auth.md", rounds: 3)
├ Round 1 (Gemini 2.5 Pro)... ✓1 critical, 0 advisory
├ Round 2 (GPT-5.2)... ✓0 critical, 2 advisory
└ Round 3 (Grok 4)... ✓0 critical, 1 advisory (consensus)
● Review complete. Here are the findings:
criticalMigration drops active sessions without grace period.
advisoryAdd rollback trigger on error rate spike.
● The critical finding about session handling is worth addressing. Let me update the plan...

Sound familiar?

It's 3am, deadline looming, weekly limit at 85%. Plan mode after plan mode and still debugging. Context bloats and dynamic limits become even more unforgiving.

Rings a bell?

Plug the Reviewer in and get your Claude a fresh pair of eyes. Ground it. Save your night for peanuts per run.

The edge in your back pocket

Catch gaps before implementation

A plan that looks solid to Claude may have false assumptions or edge cases that other models will catch. Proven. One pass, multiple perspectives.

Your subscription, stretched

Every failed implementation burns tokens on debugging and rewriting. Catching issues at the plan stage costs a fraction. Hit 85% of your weekly limit? 1K4 picks up the slack.

Zero workflow change

MCP integration. Claude Code calls the tool automatically when you're in plan mode. You write plans the same way you always do.

Evidence, not opinions

Every finding references your actual code. Not generic advice. Reviewers see your project files, your imports, your data structures.

One command. Every project.

Global MCP registration. Works everywhere Claude Code runs.

$ npx -y -p @1key4ai/mcp-review@latest mcp-review setup --setup-token YOUR_SETUP_TOKEN

Placeholder shown. Real setup token issued after sign-in on the install page (P0 security gate).

Start setup

Pricing

Free

Your first review

Flagship models. 3 rounds.
No credit card.

Per token

After that

~$1 per review on average.
Credits never expire. Budget caps.