AI can produce critical code faster than teams develop the understanding to maintain it. Ninchi shows where understanding has been demonstrated — and where it hasn't — through lightweight checks in your Git workflow.
Free to start · No credit card required
When implementation accelerates faster than shared understanding, teams inherit critical code that fewer people can explain, review, or maintain.
“Today's shortcut can become tomorrow's operational dependency.”
Ninchi Shows Where Understanding Has Been Demonstrated
Connect a repository once. After that, each change becomes one focused challenge — under a minute — and the evidence builds into team views. Teach Me steps in only when an answer misses something.
Install Ninchi for selected repositories through the GitHub App. GitLab is supported too; this clip shows the GitHub flow.
Ninchi reads the change and asks one focused question about it.
A pull-request comment opens one focused browser challenge, designed to take less than 60 seconds. Ninchi records the timestamped result as evidence in personal and organization views.
When an answer misses something, Teach Me turns that gap into a short lesson on your change.
A follow-up question on the same area records whether the gap has closed.
Recorded challenge events roll into organization analytics, knowledge maps, and accountability views. These are deterministic projections of stored evidence — not predictions of ability.
For Engineering Teams
For Educators & Training Programs
For Developers, Students & Vibe Coders
For Governance & Audit
Install Ninchi through GitHub Marketplace, or procure Enterprise via AWS Marketplace.
GitLab
Vanta
SOC 2 Type 1Ninchi ships official GitHub and GitLab integrations, live today. We partner with Vanta and Drata through their official partner programs, and Ninchi has completed a SOC 2 Type 1 attestation.
Teams and organizations using Ninchi today.
See how teams and individuals use Ninchi to make understanding visible during AI-assisted work—without turning early evidence into claims it cannot support.
Experienced engineering team
Eden is testing Ninchi inside a real engineering workflow. The case study focuses on whether lightweight verification can preserve developer engagement without disrupting delivery.
Public findings will be limited to results Eden and Ninchi have reviewed together, with measured outcomes added only when the data is ready.
Early-career engineering team
A team of interns used Ninchi in blocking mode while building a working relationship-management product with AI-assisted development tools.
Across 87 scored challenges, the team recorded an approximately 92% pass rate. Three survey respondents consistently described Ninchi as supporting better codebase understanding.
The survey sample was three interns and should be read as qualitative evidence, not a controlled study.
Education pilot · Internal evaluation
Instructors and staff at the Tokyo coding bootcamp are evaluating Ninchi through GitHub as they prepare for possible future classroom use.
This is an internal pilot in progress. Students are not using Ninchi today, and no learner outcomes or cohort metrics are being claimed.
Individual learning record
Ikye, a member of North Texas Mensa, recorded eight scored PR challenges. The only failures occurred on the initial generated scaffold, and the next six challenges were consecutive passes as the work moved from accepted boilerplate toward code Ikye could explain.
8 scored challenges · 6 passes · 6 consecutive passes after the initial failures
One person and eight challenges are a small sample. This is an observed sequence—not evidence that Ninchi caused improvement or a measure of permanent ability.
The Ninchi Score is a transparent, difficulty-weighted evidence ratio over verified-understanding events. It records what a developer or team has demonstrated in Ninchi challenges, with artifact context and timestamps, so the conversation stays grounded in inspectable evidence.
“Demonstrated understanding and accountability, made visible as evidence you can inspect.”
Ninchi combines repository facts with recorded challenge evidence to show which areas matter most and where your team has evidence on record. Every view comes from stored events and deterministic rollups; none predicts ability.
Available on the Enterprise plan.
Ninchi scans a repository and organises it into key areas ranked by size, recent activity, and cross-area dependency centrality. It also shows the repository's language mix and refreshes as the codebase changes.
Platform leads use the ranking to scope onboarding, audits, and pilots around the code that actually matters.
Each key area shows how many contributors have qualifying evidence: passing challenge answers tied to that part of the repository. Ninchi weights each area's coverage classification by importance, so higher-importance areas contribute more to the overall figure. Areas with one qualifying contributor are labelled single-contributor.
Understanding Debt is the share of key-area importance with no verified-understanding coverage. Ninchi calculates it from recorded challenge evidence.
Single-contributor areas are key-person risk you can see before a departure makes it urgent — and Understanding Debt is one deterministic number a quarterly review can track.
Org admins can inspect one developer's recorded evidence by repository area. The view separates evidence inferred from PR challenge history from evidence gathered through targeted assessments. It reports coverage and does not rank people.
For onboarding and handovers: see where understanding has been demonstrated — not just who touched what — without ranking people.
Available now is supported today. In development is active work not yet available. Exploring is directional and not committed.
GitHub and GitLab integrations create challenges from code changes inside the review cycle.
Early prototype for creating a challenge from manually submitted plain-text or code excerpts without connecting a repository.
Potential challenge-and-evidence flows for AI-assisted drafts, memos, and reports.
Add lightweight verification without replacing the review process your team already uses.
Tie each recorded check to the person, artifact, and point-in-time decision it covers.
Detailed records of human understanding and approval at every step.
Review aggregate evidence patterns across teams, repositories, and knowledge-map areas.
Every verification records the question, answer, score, difficulty, artifact context, and timestamp.
“Keep AI-assisted delivery fast and human ownership visible.”
Unlimited members on every plan. Start with the core challenge loop, then add team analytics, stricter verification, and enterprise controls.
Swipe to compare all four plans.
| Compare plans | Enterprise Custom tailored to your organization Organization-wide evidence, exports, knowledge maps, and support. Contact Us | |||
|---|---|---|---|---|
| Core workflow | ||||
| Scored PRs | 5 / month (soft cap) | Unlimited | Unlimited | Unlimited |
| Members | Unlimited | Unlimited | Unlimited | Unlimited |
| Verification modes | Casual | Casual, Tracking, Blocking | Casual, Tracking, Blocking, Strict | Casual, Tracking, Blocking, Strict |
| Question difficulty | Easy | Easy, Medium | Easy, Medium, Hard | Easy, Medium, Hard |
| Diff preview in challenges | — | |||
| Organization analytics | — | |||
| Advanced verification | ||||
| Anti-cheat controls | — | — | ||
| Teach Me lessons | — | — | Available when enabled by an org admin | Available when enabled by an org admin |
| Governance and audit trails | — | — | ||
| Enterprise capabilities | ||||
| Audit log export | — | — | — | |
| Knowledge maps and baselines | — | — | — | |
| Dedicated support and SLA | — | — | — | |
| Custom branding | — | — | — | |
| AI spend analytics | — | — | — | |
| AI spend vendor integrations | — | — | — | |
Standard and Pro prices are per seat, billed monthly. The Hobbyist scored-PR limit is a soft cap: challenges continue, but additional PRs are not scored that month.
Read the research behind Ninchi's challenge loop, inspectable evidence model, and transparent difficulty-weighted score.
The paper clearly separates shipped capabilities from proposed models and validation targets.
Straight answers about workflow impact, privacy, evidence, and the larger Ninchi vision.
Ninchi is built for software teams today. We are prototyping how the same challenge-and-evidence model could support manually submitted text in adjacent areas; broader non-software workflows are not generally available.
Ninchi was founded around a simple conviction: AI can accelerate production without turning people into passive reviewers. We build lightweight verification and inspectable evidence for teams that want speed and human ownership together.
Founder, CEO & CTO
CPO & CMO
Head of Research
Book a working session or share your source platforms, team size, and security requirements. We will route the right pilot and onboarding path.
Book Team DemoAffiliate, design-partner, or investor inquiry? support@ninchi.ai
New features, product insights, and early access — no spam, unsubscribe any time.