Sensei — AI Code Review & Comprehension

Why Sensei

Code is now written faster than anyone can read it. Agents open pull requests while you sleep, and the diff is always longer than your attention. Sensei does not write your code for you. It reads it with you, line by line, until you understand it, and tells you honestly how good it is.

Watch a review

Five snippets. Four answers each.

Code an AI could have written this morning. Pick a language, then read what Sensei says about it: the explanation, the issues, the grade and the fixes.

auth/user-auth.tsTypeScript, 9 lines

1export async function authenticateUser(email: string, password: string) {
2 const user = await db.query(3
3 `SELECT * FROM users WHERE email = '${email}'`1
4 )
5 if (password === user.password) {2
6 return createSession(user)
7 }
8 return null
9}

Looks up a user by email and signs them in when the password matches. It builds the SQL by hand from the email and compares the password as plain text, so the login can be bypassed, and a missing user crashes the request.

How it works

One beam in. Four answers out.

Paste a snippet, drop a file or point at a GitHub link. Sensei detects the language, reads the whole thing and splits what it finds into four answers.

Explains
Flags
Grades
Fixes

01Explain

Plain English, not a wall of jargon.

One clear paragraph per function: what it does, what it touches and where the sharp edges are. Then ask follow-ups until it clicks.

`SELECT * FROM users WHERE email = '${email}'`

Finds the user by email, but builds the SQL by hand from the email, so a crafted email can sign in as anyone.

Why is line 3 dangerous?

Because the email goes into the SQL string as-is. Someone can sign in as anyone by typing ' OR '1'='1 as their email. Pass it as a parameter instead and the database treats it as data, never as SQL.

02Flag

Every risky line, ranked.

Security holes, smells and needless complexity, each pinned to its line and sorted from critical to low, with the time the fix should take.

1criticalSQL injectionauth/user-auth.ts:35 min
2criticalConcurrent map writesjobs/worker.go:520 min
3highPlaintext passwordauth/user-auth.ts:530 min
4highMutable default argumentorders/cart.py:110 min
5highIgnored errorjobs/worker.go:410 min
6highIndex cannot be usedreports/revenue.sql:45 min

03Grade

A grade you can check by hand.

Three sub-scores and the formula that joins them. Move them and watch the grade change. 90 and up is an A+, anything under 50 an F.

Health score

70/ 100

Good

Complexity32Lower is better: the branching and nesting a reader has to hold in their head.Technical debt45Lower is better: smells, anti-patterns and the refactoring they will cost.Maintainability88Higher is better: naming, structure and documentation for the next reader.

( 88 + (100 − 45) + (100 − 32) ) ÷ 3 = 70

04Fix

Prove the refactor worked.

Re-run Sensei on the same file and compare the two reviews: the score change, the issues resolved and any new ones. Drag the line to see this one.

Before · 42F

1export async function authenticateUser(email: string, password: string) {
2 const user = await db.query(
3 `SELECT * FROM users WHERE email = '${email}'`
4 )
5 if (password === user.password) {
6 return createSession(user)
7 }
8 return null
9}

After · 87A

1export async function authenticateUser(email: string, password: string) {
2 const user = await db.query('SELECT id, hash FROM users WHERE email = $1', [email])
3 if (!user) return null
4 const ok = await verify(user.hash, password)
5 return ok ? createSession(user) : null
6}

1export async function authenticateUser(email: string, password: string) {
2 const user = await db.query(
3 `SELECT * FROM users WHERE email = '${email}'`
4 )
5 if (password === user.password) {
6 return createSession(user)
7 }
8 return null
9}

FBefore · 42After · 87A

42 → 87+45SQL injection patchedPassword hashing addedMissing user handled

Your workspace

Every review, kept and shareable.

Sensei's glass prism on a desk at night, splitting a beam into four colored rays — Reviews
Every file you review stays in your workspace, with its explanation, issues, grade and fixes.

History

Watch a file get better.

Every analysis is saved with its grade, so you can re-open it, compare it with a later run, or export it.

auth/user-auth.tsToday · 87 / 100Grade A
orders/cart.pyYesterday · 61 / 100Grade C
src/config.rsMon · 75 / 100Grade B

One link, read-only.

Send a review to a teammate or paste it into a pull request.

sensei.jishulabs.com/s/7Kq2…Copy

The explanation
Every issue, by line
The grade and the suggested fixes

Languages

Fifteen-plus languages.

TypeScript · JavaScript · Python · Go · Rust · Java · Kotlin · Swift · C# · C++ · Ruby · PHP · SQL · and more

Private by default

Your code stays yours.

Code and reports are scoped to your account, encrypted at rest and never used to train models. Export everything as JSON, or delete it, whenever you like.

Pricing

Start free. Upgrade when it sticks.

One Pro plan for every working developer. Teams that need single sign-on, shared standards or self-hosting talk to us.

Free
For trying Sensei on real code
$0forever
- 5 analyses a month
- Every supported language
- Explanations, issues and grades
- Full history
Start free
Pro
Recommended
For developers who review code daily
$15per month
Billed annually, or $19 month to month.
- Unlimited analyses
- Priority processing
- Advanced security scanning
- Compare any two analyses
- Export reports
- Priority support
Start with Pro
Enterprise
For organizations with security requirements
Custom
- Everything in Pro
- Single sign-on
- Self-hosting option
- Shared review standards
- Dedicated support
Talk to us

Questions

Asked and answered.

Anything else? Ask us directly.

How is Sensei different from a general AI chatbot?

Sensei is purpose-built for code comprehension and review. It explains code in plain English and runs a structured, multi-dimensional review with quality and complexity scores and specific improvements, rather than a free-form chat. Think of it as a senior reviewer focused on your code.

How accurate and reliable is the review?

Sensei gives consistent, multi-dimensional feedback with concrete suggestions and clear scoring. It is a strong, always-available second opinion, and you stay the decision maker. Review the suggestions, apply what fits, and keep the context only you have.

Can I act on or edit the suggestions?

Yes. Every explanation and suggestion is meant to be applied directly to your code. You decide which improvements to make, and you can re-run Sensei afterward to confirm the quality and complexity scores moved in the right direction.

Is my code private?

Your code and reports are scoped to your account, encrypted at rest, and never used to train models. Enterprise plans add a self-host option so your code never leaves your environment.

How does pricing work?

Free includes a generous number of analyses with explanations and review, no credit card required. Pro is $15 per month for unlimited analyses, scoring, maintainability analysis, report export, and history. Enterprise is custom with SSO, self-hosting, and dedicated support.

Which languages does Sensei read?

Fifteen-plus, including TypeScript, JavaScript, Python, Go, Rust, Java, Kotlin, Swift, C#, C++, Ruby, PHP and SQL. The language is detected for you and the feedback follows each language’s idioms.

How is the grade calculated?

Three sub-scores, maintainability, complexity and technical debt, fold into one health score: (maintainability + (100 − debt) + (100 − complexity)) / 3. 90 and up is an A+, 80 an A, 70 a B, 60 a C, 50 a D and anything lower an F. The math is shown on every analysis.

Can I prove a refactor actually helped?

Yes. Re-run Sensei on the same file and compare the two analyses: score deltas, resolved issues, new issues, and functions added or removed.

Sensei's glass prism splitting a white beam into four colored rays

Understand every lineyou ship.

Paste a file and get a grade, a plain-English explanation and a ranked list of fixes. Free for five analyses a month.

Start free Talk to us

Every line,broken down.