Connect the repository
Pick the repo and branch. VrittOS counts the source files — not fixtures, lockfiles or vendored code — and quotes the scan before anything is created. Asking the price costs nothing.
For teams with a product already in GitHub
Add features to the codebase you already have
Connect the repository. VrittOS reads the code, writes the requirements, architecture and backlog it implies, and from then on plans every feature you add against that code — delivered as a tested pull request on a branch, never to main.
Founding offer 50% off your first 3 months — ends 31 October 2026
✓ Scan quoted before it starts — 20, 40 or 80 credits by repo size ✓ You review what it found before anything is planned on it ✓ Adding a feature never uses a project slot ✓ Feature branches and pull requests only
From an existing repo to the next pull request
It reads before it writes
The scanner maps the codebase, works out the stack, identifies the modules, and writes an as-built PRD, an architecture document and — if there is a UI — the design system. Usually three to six minutes, narrated step by step.
You correct it
Inference from code is wrong in places. The scan lands at a review gate where you edit the requirements before approving, so nothing is ever planned against a misunderstanding. If the scan fails, the credits come back.
Add the feature
Describe it. The planner asks only what it cannot infer from your code and prior requirements, writes stories against your existing modules, and the developer agents open tested pull requests on feature branches.
Built for code that already has opinions
A codebase with a year of decisions in it deserves an AI that reads them before adding to them. That is what the scan is for.
Priced by size, not by surprise
Up to 300 source files is a small scan, up to 1,500 medium, beyond that large. The price buys depth of reading, it is fixed before you start, and it is refunded if the scan fails.
Your modules, your stack
The tech stack is inferred from your manifests and fixed for the project. New stories name the modules they touch, so a feature extends your services rather than inventing parallel ones.
Your workflow stays yours
Work lands on feature branches as pull requests with test results in the body. Nothing is pushed to main. CI results sync back; a red build on main gets triaged before new work proceeds.
Wrong inference is cheap to fix
The as-built requirements are an editable document at the review gate. Fix the line that is wrong once, and every feature planned afterwards inherits the correction.
The design system is captured too
For repositories with a front end, the scan records your theme, layout shell and component conventions, so generated screens match the ones you already ship.
The map is rebuilt every time
The developer agents rebuild their view of the code at implementation time, so changes your team merged outside VrittOS are seen. Only the requirements and backlog are snapshots — and you own those.
Common Questions
What does the scan cost?
It depends on how many source files the repository has: 20 credits up to 300 files, 40 up to 1,500, 80 beyond that. The quote is shown before you start, charged when you start, and refunded automatically if the scan fails. It is available on every plan, including pay as you go.
How long does it take?
Usually three to six minutes. The scan narrates each stage — cloning, mapping the codebase, working out the stack, identifying modules, writing the requirements and architecture, reconstructing the backlog — so you can see it working rather than wonder whether it is stuck.
What if it gets my architecture wrong?
That is what the review gate is for. The scan stops and shows you what it found; you edit the requirements document before approving. Every feature added later is planned against the corrected version, not the guess.
Will it ever push to main?
No. Every change is a feature branch and a pull request. Merging is your decision, made in GitHub, with the tests and CI results in front of you.
Which languages and stacks does it understand?
It recognises the common web and API stacks — .NET, Node and TypeScript, Python, Java, Go, Ruby, PHP — with Angular, React, Vue and Next.js front ends, and Flutter for mobile. Anything else is scanned with a free-form stack description; you can correct it at the review gate.
Does it work on private repositories and monorepos?
Yes. You connect GitHub through OAuth and the token is stored encrypted; the scanner reads only the repository and branch you choose. Monorepos are sampled across their packages so no single one crowds the others out.
Build one real feature.
Then judge us on the pull request.
Bring a feature from your existing backlog. A team of specialised AI agents takes it through requirements, stories, code and tests — you keep every decision that matters.
- 18
- Specialised AI agents
- Hours
- From idea to first PR
- 4
- Human approval gates
- £29
- Plans from, per month · or pay as you go
Ready to transform your software delivery?
14-day free trial · 100 credits · No credit card required