From a recorded Claude Code build

Claude Code missed 11 issues in its own plan. Here’s how you catch them in yours.

No code to write. Works inside the Claude Code or Cursor chat you already use.

Free account. Verify a card for $0.50 and get $5 credit to review with.

What the independent reviewers caught

Claude Code reads findings like these and decides what to fix. The small line on each card is the finding exactly as it was recorded.

CriticalOperational
The spending limit could not actually stop spending.
As recorded: No actual USD cost parsing from response.usage — cost cap is unenforceable
Caught by: MegaLens (all engines)
CriticalOperational
Losing its database would make it redo drafts for old emails, in bulk.
As recorded: Bootstrap is :unread query + no gmail.modify — DB loss causes mass re-drafting of old mail
Caught by: MegaLens (judges)
HighSecurity
An email's own subject line could slip instructions to the AI.
As recorded: Email headers (subject, from, date) injected raw into LLM prompt — no sanitization
Caught by: Both (pre-audit missed headers)
HighOperational
A crash at the wrong moment would create duplicate drafts.
As recorded: Draft-create → mark-processed crash window — duplicate drafts, no atomic guard
Caught by: MegaLens (all engines)

From a recorded build. Its three reviews cost $3.32. Full case study here.

Watch the recorded build · 1:09 · no sound

A sped-up screen recording of the build. The clip runs 1 minute 9 seconds with no sound; the build itself took a lot longer. The findings tables are held on screen so you can read them.

Three steps, about two minutes

Create a free account

Connect it to your coding tool

One command, pasted into the terminal on the computer where you use Claude Code. This step needs that computer, not your phone.

Paste into your terminal

npx megalens-mcp setup

It finds Claude Code, asks for your connection token, which you copy from your MegaLens settings, and does the setup for you. Then restart Claude Code and type run megalens_status in the chat. A reply means you are connected.

Using Cursor or Codex? Each has its own short setup guide: Cursor · Codex

Ask for a review in plain English

Type either of these into your chat. Your coding tool takes it from there.

“Review this plan with MegaLens before we build”

“Run a MegaLens security audit on what we just built”

Example of what comes back in your chat

> Review this plan with MegaLens before we build

megalens_debate · done

Independent reviewers returned findings, including:

[Critical] No actual USD cost parsing from response.usage — cost cap is unenforceable

[High] Email headers (subject, from, date) injected raw into LLM prompt — no sanitization

I checked these against the plan. Both hold up. Updating the plan before we build.

Shown as text, not a screenshot. The two findings are real, from the recorded build.

What happens next

The findings come back into your chat.
Claude Code reads them and decides what to fix.
You stay in the same conversation.

You never leave your IDE. You never read a report.

Findings like these, on your own project, in the chat you already use.

Try it free on your project

$5 free credit covers a build like this one. No subscription.

Questions

Do I need to know how to code?

No. You run one setup command, then ask for reviews in plain English inside the chat you already use. Your coding tool reads the findings and handles the fixes.

What is an MCP?

A small add-on that lets your coding tool talk to an outside service. The setup command in step 2 installs it for you, and after that you just type normal English.

How can I try it?

Creating an account is free and needs no card. Verify a card for $0.50 and you get $5 in credit, enough for a full build’s reviews. Or connect your own OpenRouter key and review with no MegaLens charges at all.

Will it send my code somewhere?

Yes. The code you ask about is sent to the AI models that review it. Your review history is kept for 30 days on the free plan and 90 days on paid plans, then deleted automatically. Details are in the privacy policy.

Does it replace Claude Code?

No. It adds independent reviewers to it. Claude Code still writes the code and still decides which findings to act on.

What does it cost after the free credit?

Two ways to pay, no subscription needed for either.

Pay as you go: you keep a balance with MegaLens and each review is charged from it. In the recorded build, the plan review cost $1.25, the mid-build code review $1.23, and the final security review $0.84, so $3.32 for the whole build. Your $5 starting credit covers a build like that.

Bring your own key: connect your OpenRouter key and pay the AI providers directly, roughly $1.50 to $2 for the same three reviews, with two reviewers plus a judge instead of three. MegaLens charges nothing on this path.

Pro at $15 a month adds a deeper review with five models per run, live-data specialists like Perplexity and Grok, automatic fix routing in Claude Code and Codex, and no daily limit on reviews. Full details on the pricing page.

Want to see it for yourself? There is a screen recording of the whole build, every finding, and what Claude rejected. Watch the recorded build