Insights
How long does it take to build an MVP or a custom platform?
By Rare Agentics engineering teamLast updated First published 7 minute read
With Rare Agentics a single feature lands in 1 to 2 weeks, a first product typically takes 1 to 8 weeks, and larger platforms are delivered in phases of 6 to 8 weeks, each priced before it starts. Before any of that, you get a first view within two working days and a written plan, a fixed price and the risks within a week. What decides where you land in those ranges is mostly scope and how quickly decisions are made, not typing speed.
How long does each kind of build take?
Timelines only mean something when you know what is being built. These are the ranges we work to, and the shape of each.
| What you are building | Typical length | What you have at the end |
|---|---|---|
| Scoping | Up to a week, free | A written plan, a fixed price and the risks |
| A single feature on a running product | 1 to 2 weeks | One feature live in your accounts |
| A proof sprint | 1 to 2 weeks | One bounded feature, workflow or working prototype, live |
| A first product | 1 to 8 weeks | A product real users can use, in production |
| A custom AI agent for one task | 3 to 6 weeks | An agent working within written bounds, every action logged |
| A platform | Phases of 6 to 8 weeks each | Each phase live and tested before the next is priced |
The spread inside each range is real. A first product with a handful of journeys, one kind of user and no integrations sits near the start of it. One with several roles, payments, an accounting connection and personal data sits near the end. The same drivers set the price; see how much an MVP costs in the UK.
What happens before the build starts?
The fastest projects are the ones where the first week is spent agreeing what to build, not building the wrong thing quickly.
Day 1 to 2. You describe the outcome you want in plain language. Within two working days you get a first view: whether we think it can be done, roughly how, and what we need to know.
By the end of week 1. A written plan, a fixed price and the risks. The plan lists the journeys in scope, the ones out of scope, the acceptance criteria and the order of work.
Approval. You approve the plan and the price before the build starts. Nothing is built on an assumption you have not seen.
Scoping a build you already have in mind is free. If you are not yet sure what to build, a short discovery may be the better first step; we explain when in what a discovery phase is, and whether you need one.
What does a phase-by-phase calendar look like?
Here is the shape of an illustrative first product of around six weeks. Yours will differ, but the rhythm is the same: something working every week, and nothing released without a person signing it off.
| Week | What happens | What you see |
|---|---|---|
| 0 | Scoping, plan, price, approval | The written plan and fixed price |
| 1 | Foundations: accounts, environments, sign-in, data model | A staging site you can log in to |
| 2 | The first core journeys | The main path working end to end |
| 3 | The remaining journeys and roles | Each user type seeing what it should |
| 4 | Integrations: payments, email, outside systems | Real connections in staging |
| 5 | Testing, security checks, edge cases, admin screens | Fixes, and the tests behind them |
| 6 | Release to production, handover | The product live in your accounts |
A platform follows the same rhythm, one phase at a time. Each phase of 6 to 8 weeks has its own plan and fixed price, agreed before it starts, and is tested by your leadership before it goes wider. You can stop after any phase and keep everything built so far. Our platform build page sets out how phases are chosen.
What makes software timelines slip?
Projects rarely run late because the engineering was slow. They run late for reasons that are visible early if you look for them.
- Decisions that wait. A question about pricing rules or who may approve what, left open for a week, holds up everything that depends on it.
- Scope that grows mid-build. Each "while you are at it" is reasonable on its own. Together they are a second project.
- Integrations that surprise. An outside system with poor documentation, rate limits or no test environment can take longer than the rest of the build.
- Content and data that are not ready. Copy, prices, product data and the spreadsheet you want migrated all have to arrive on time.
- Access that takes weeks. Accounts, domains and permissions that only one busy person can grant.
- Approval queues. App store reviews, a partner's security questionnaire or a board sign-off, each with its own calendar.
None of these is unusual, and a good plan names them as risks in the first week rather than discovering them in the fifth.
How do AI agents change the timeline, and what do they not change?
We work as a small team of senior engineers with AI agents doing the volume of the work. Agents build in parallel, each in its own separate workspace, and they test and review their own output before a person reads every result. That shortens the parts of a build that used to be limited by how many hands were available: writing the code for a well-defined journey, writing the tests for it, and checking it against the rules.
What agents do not shorten is just as important.
- Decisions. Only you can decide how your business should work. An agent cannot tell you whether a refund needs a manager's approval.
- Approvals. You approve the plan and the price before the build starts, and a named engineer ships every release. Agents have no access to production.
- Outside calendars. App store reviews, partner integrations and your own team's availability run at their own pace.
- Learning from users. An MVP exists to learn something. The learning takes as long as it takes.
Agents do not make every project faster. With the scope and decisions settled, the build itself is rarely the slowest part any more. We explain how agents are kept away from live systems in how to keep AI agents away from production.
What can you do to ship sooner?
Most of the time you can save is on your side of the table. A short list that helps every build:
- Name one decision-maker who can answer questions within a day.
- Write down the one question the first version must answer, and cut anything that does not serve it.
- Start with a responsive web app unless native mobile features are essential.
- Grant access early: cloud accounts, domains, the outside systems you want connected.
- Get content and data ready before the week they are needed.
- Keep a "later" list for good ideas that arrive mid-build, and price them after launch.
- Look at working software every week, and say early when something is wrong.
- Start with a proof sprint if one part of the idea is genuinely uncertain. One working thing in 1 to 2 weeks settles a debate faster than a long specification.
Why do quoted timelines vary so much between suppliers?
If you ask three suppliers how long your product will take, you may get three very different answers. Usually they are not answering the same question.
- They are quoting different scopes. One has priced the five journeys you described; another has quietly added the admin screens, the reporting and the mobile app they assume you will want. Ask each to list what is in and what is out.
- They count different things. Some quote only build time. Others include scoping, testing, release and handover. Ask when the clock starts and what "done" means.
- They staff differently. A large team with juniors doing the volume spends time coordinating that a small senior team does not, though very large projects can need the extra hands.
- Some are guessing. A timeline given before any questions about your users, systems and data is an opening bid, not a plan.
The useful comparison is not the shortest number. It is the timeline attached to a written scope, acceptance criteria and a fixed price, from someone who has named the risks. We set out what else to ask in questions to ask a software development company.
Is it better to launch sooner with less, or later with more?
For a first product, sooner with less is almost always right. Every week before launch is a week of building on assumptions; every week after it is a week of learning from real users. A smaller first release also costs less, carries fewer places for things to go wrong, and makes the second phase better informed.
The exception is anything that must be right on day one: payments, personal or health data, safeguarding, and the rules that stop one customer seeing another's records. Those are never the corners to cut for speed. The way to launch sooner is to do fewer things, not to do the same things less carefully.
How long does it take to get changes after launch?
Launch is not the end of the timeline. Every build carries the launch warranty: any defect against the agreed acceptance criteria, found in the 30 days after release, is fixed at no cost. After that you choose: a Run and improve plan with agreed response times, or a full handover with documentation your engineers can use.
Small changes on a running product usually land on the same rhythm as the build: a single feature in 1 to 2 weeks, each priced in writing before it starts.
How Rare Agentics can help
We build bespoke software and operating platforms for founders and growing businesses, with senior engineers leading every build and AI agents doing the heavy lifting. Your code and data live in your own accounts from the first commit, agents cannot reach production, and a person signs off every release.
Tell us what you need built. Within two working days you get a first view, and within a week a written plan, a fixed price and the risks.