Tools
Free tools that don’t want your email.
The convention in this industry is a calculator that collects ten answers, then asks for a work email and promises a number within a business day. That’s a lead form wearing a tool’s clothes, and the arithmetic behind it is usually a lookup table nobody will show you.
Everything here works the other way round. It runs in your browser, it shows the arithmetic, it lets you disagree with the assumptions, and it never asks who you are. If a tool on this page ever needs an account before it’s useful, it won’t go on this page.
Before you write the brief
All three run before anybody quotes anything, because the most expensive mistakes on a software project are made here: asking for the wrong thing, asking for it on the wrong platform, and asking for it in a document nobody can price.
Project brief generator
Answer a structured set of questions and read a complete, vendor-ready scope brief on the page, with every gap in your own thinking named.
Context, problem, launch scope, explicit exclusions, integrations, data and compliance constraints, platforms, users and roles, scale, timing, budget band and decision process. Alongside the document it scores completeness, says why each missing section changes a quote, lists the assumptions a vendor will otherwise have to make on your behalf, and writes the follow-up questions a competent one will send back.
Deterministic templating. No model call, no email.
App or website decision tool
Tick the device capabilities your product genuinely needs and see which build is forced, which single requirement forced it, and why the browser cannot cover it.
Four verdicts: responsive web app, installable web app, cross-platform native, or fully native. It names the requirement that decided the answer, lists the capabilities that were checked and changed nothing, and shows the full rule trace so you can argue with it. Then the part the blog posts on this question skip: what the verdict costs in store review latency and install friction, and the commission exposure if you sell digital goods.
Every rule shown, including the ones that did not fire.
Tech stack decision tool
Answer what the site has to do and see which platform your requirements force, which single rule forced it, and what that answer costs to run and to leave.
Twelve questions covering who edits content, the content model, accounts and permissions, what is sold, integrations, compliance constraints, traffic, design and who maintains it afterwards. Six verdicts: a site builder, WordPress with a theme, Shopify with a theme, Shopify behind a custom front end, a headless CMS behind a custom front end, or a full custom application. Then the part the comparison posts leave out: recurring plan fees taken from the vendors' own published pricing, the cost of migrating off later, who you can hire to maintain it, and the lock-in that is not obvious on day one.
All thirty rules shown, including the nine that raise nothing.
Before you agree a number
One prices the build, one prices the thing that keeps costing money after launch, and one prices the manual process you are trying to stop paying for. All three show every line, and all three let you disagree with the assumptions rather than take a total on trust.
MVP cost estimator
Pick a platform and the feature areas you need, and watch a costed line-item breakdown build itself as you type.
Discovery, design, build broken out by feature, integrations, QA, environment setup and deployment, each as an hour band you can edit. Then the three costs that fall outside almost every build quote: third-party running costs, store fees, and first-year maintenance. The output is a range, and the page explains what widens it.
No email. No signup. Nothing leaves your browser.
AI feature cost per user calculator
Work out what one user of your AI feature costs a month, then divide it into what you charge and read the gross margin.
Takes a usage shape rather than a token count: active users, sessions, model calls, and a per-call token budget split into system prompt, retrieved context, replayed history and output. Then it applies cached-input pricing on the repeated prefix, a retry rate, and an optional cheap-model routing split. Ends where the token calculators stop: margin against your subscription price, and the price at which the feature breaks even.
Dated, source-linked provider list prices. No email.
Automation ROI calculator
Price what a manual process costs you a year, then read how long an automation takes to pay for itself once you count what it costs to own.
Annual manual cost split into labour hours and rework, from a named task, headcount, frequency, minutes per run, loaded hourly cost, an error rate and the cost of a mistake. Against it, the four running costs an ROI post leaves out: the runs the automation cannot cover, the human reviewing exceptions, model and vendor spend, and maintenance. Payback is net of all four, in months and in runs, alongside the largest build cost that still lands inside your own horizon. Then nine published checks that say when not to believe the payback figure.
Nine checks shown, including the ones that did not fire. No email.
Before someone else audits you
Two examinations: the one you run on yourself the week before you go live, and the one an acquirer, an investor or an incoming agency runs on you later. Both checklists are public. What neither of them publishes is the weighting, which is the part that decides whether a finding blocks a launch, reprices a deal, or gets waved through.
Launch readiness scorer
Score what is genuinely ready in the last two weeks before you go live, then find out whether the work that actually blocks a launch fits in the days you have left.
Twenty-three items across security, data, reliability, performance, legal and compliance, and operations, each weighted in the open. Four of them are treated as launch blockers rather than points, and any one left open caps the whole score: a backup nobody has restored, secrets in the repository, no error tracking, and a rollback nobody has rehearsed. The output is a remediation order sorted by risk cleared per engineer-day, a plain split of what blocks a launch and what can honestly follow in week two, and the plan measured against the working days you have rather than the ones you wish you had.
Four blockers cap the score rather than average away. No email.
Technical due diligence scorer
Answer thirty-five questions across six weighted areas and read the score a reviewer would arrive at before they arrive at it.
Key-person risk, deployment and environments, test coverage on revenue paths, security and secrets handling, licensing and the IP chain, and data handling checked against your own published privacy policy. Findings are ranked by how reliably each one gets a deal repriced rather than by generic severity, with a separate list of the ones that look alarming on a checklist but rarely move a price, and a remediation order sorted by price protected per engineer-day.
The weighting is shown. Nothing is uploaded.
Eight of them so far, and each one is here because the same question kept arriving by email until automating the answer was less work than retyping it. That’s the only thing that gets one built, so the next one arrives when the next question does rather than on a content calendar.
Past the estimate stage?
Tell us what you’re building and what it has to do by when. Our posted price bands sit on the services pages, so you can check whether we’re in your range before you spend an email on us. Those bands are estimates in US dollars, not quotations; the quote that follows scoping is fixed in writing before kickoff.