REQ-4128Password self-reset · now in AI Review

From feature request to shipped — with a human in the loop.

ZenBuild gives every feature one calm, reviewable path: AI drafts the PRD, plans the tasks, and reviews the pull request against your requirements — but a person always approves the release.

No credit card · Connect a GitHub repo in minutes

How it works

Five phases. One clear path to production.

Request → PRD → Tasks → Code → AI Review → Fixes → Approval → Ship. Every stage is visible, and nothing reaches production on the AI's word alone.

01Discovery

Understand the request — and push back when needed

A request arrives by email, ticket, or call. The AI agent gathers missing context with follow-up questions, and tells you when the capability already exists. Only what's truly needed moves forward — then it drafts a structured PRD.

Clarifying QsPRD draftAcceptance criteria
02Planning

Turn the PRD into tracked engineering work

ZenBuild breaks the spec into actionable tasks, organized on a Kanban board. Your team reviews and approves the plan before a single line is written.

Task breakdownKanban boardTeam approval
03Development

Connect the repo, open the pull request

Link a GitHub repository and let developers — or coding agents — implement the feature. Pull requests carry the real changes that the spec called for.

GitHub via OctokitWebhooksReal diffs
04AI Review

Review the code against the requirements, not just syntax

The QA agent checks the PR against the PRD, acceptance criteria, security, performance and edge cases. Issues are sorted into blocking and non-blocking — each with a reason. Fixes loop back and re-review until it's ready.

Blocking vs non-blockingRe-review loopExplained findings
05Release

A human makes the final call

A reviewer verifies the PRD, tasks, pull request and full AI review history, then approves or rejects. Only approved features move to shipped — with the whole trail preserved.

Human approvalAudit trailShipped
The platform

Everything the lifecycle needs.

One multi-tenant workspace for discovery, planning, development, AI review and release — built on tRPC, Inngest and the AI SDK.

AI review that reasons about requirements

Your QA agent evaluates whether the implementation actually satisfies the PRD — checking acceptance criteria, security, performance and edge cases, then explaining why each issue matters. Fixes loop back until it's production-ready.

1 blocking · 1 non-blocking · re-review queued

Product discovery

Reads the request, asks for missing context, and flags when it already exists.

Structured PRDs

Problem, goals, non-goals, user stories, acceptance criteria, edge cases, metrics.

GitHub, connected

Repos, webhooks, real PRs and diffs via Octokit. No hardcoded data, ever.

Tasks on a board

The PRD becomes tracked engineering tasks your team approves before development.

Human-approved ship

A reviewer signs off the PRD, PR and AI history. Only then does it ship.

Inside a review

The spec, marked up by your QA agent.

ZenBuild reads the pull request against the PRD and leaves notes in the margin — exactly where a requirement is met, missed, or at risk.

PRD-4128 · password-self-reset.md

Password self-reset

v2 · generated & human-edited · 7 files in PR #212

Problem

Users who forget their password must email support and wait hours — this drives ~18% of all tickets and blocks sign-in entirely.

Goals
  • Request a reset link directly from the login page.
  • Reduce password-related tickets by at least 60%.
Acceptance criteria
  • Reset links are single-use and expire after 30 minutes.
  • Reset emails are rate-limited to 3 per hour per account.
  • Passwords are re-hashed with the current cost factor on reset.
AI review · PR #212
BlockingTokens never expire

Validation checks the token value but not its age — the 30-minute rule isn't enforced. Real account-takeover risk if a link leaks.

auth/reset.ts:42
Non-blockingRate limit not wired up

The 3/hour limit is defined but not applied to the request route. Recommended before launch.

api/forgot.ts:18
PassedRe-hashing is correct

Uses the current cost factor — matches the criteria.

auth/hash.ts
Pricing

Start free. Scale when you ship more.

Usage-based AI credits and repository limits — billed securely through Razorpay. Change plans anytime.

Free

For solo builders and trying ZenBuild end to end.

₹0/ forever
  • 1 workspace · up to 3 members
  • 1 connected GitHub repository
  • 25 AI credits / month
  • PRD + task generation, AI code review
  • Human approval gate & full audit trail
Start free
Most popular

Pro

For individuals shipping serious side projects.

₹999/ month
  • Personal workspace
  • Up to 5 connected repositories
  • 200 AI credits / month
  • AI release-readiness assessment
  • Priority support
Upgrade to Pro

Team

For teams shipping features every week.

₹2,499/ month
  • Unlimited workspace members
  • Up to 25 connected repositories
  • 500 AI credits / month
  • Re-review loops & readiness checks
  • Webhook automation via Inngest
  • Priority support
Choose Team

Individuals pick Free or Pro; teams pick Free or Team. Need SSO, audit exports or a higher repo limit? Enterprise plans are available — talk to us.

ZenBuild

Bring calm to how you ship.

From the first feature request to a human-approved release — ZenBuild gives your team one clear, reviewable path to production.