Oktopeak

[ CUSTOM SAAS · LEGAL TECH · HEALTHCARE ]

You know the method. We build the platform your clients pay for.

We build custom SaaS platforms for founders in legal tech and healthcare. TypeScript on Next.js and Postgres, Stripe billing, documents kept in your own cloud account. Fixed scope and a date in writing. The code and every account are in your name from the first milestone.

Scope my platform with a founder

Free scoping call with a founder. You keep the written scope either way.

Legal document search · demo data
Search results screen of a legal document search platform, filters by practice area and jurisdiction
Case platform · broker view · demo data
Assigned cases table of a case management platform with status, claim type and dates
AI learning platform · demo data
Learner home of an AI learning platform: upload a document to generate a quiz, curated quizzes by level, credits and leaderboard
10 weeks
document search platform for a 17-person firm
8 weeks
HIPAA case platform, three portals
12 weeks
AI learning platform with Stripe billing
A lawyer in a dark suit writing a method out by hand on a legal pad, law books behind

Who hires a SaaS development agency for a regulated niche?

People who already do the work by hand and know exactly what the software has to do.

This is for you if

  • You are a lawyer, clinician or operator with a working method, and you want to sell it as software to others in your field.
  • Your company needs a platform of its own and renting one does not fit how you work.
  • You have a prototype, and it now has to hold real client documents, take payments and survive a security review.

Start elsewhere if

  • You want to replace Clio, MyCase or Filevine for your own firm: that engagement has its own shape at Say Goodbye.
  • You want Claude connected to the system you already run: start with Guided MCP Setup.
  • A no-code tool already does the job and the data is not regulated. Keep it.

What is in your build?

Tick what your platform needs. The weeks come from the shape of builds we have shipped, so you can see what you are asking for before anyone quotes it.

Build sheetBase build

First version, live in

5 to 6 weeks

Fast delivery and quality at the same time. Finally possible, because we are AI-native software engineers.

In itAccounts, roles, admin screen, deploy pipeline

Yours day oneGitHub organisation, Vercel or AWS account, Stripe, Clerk, domain

Scope my platform with a founder

Weeks are the shape of past builds, not a quote. Your scope on the call turns this into a date in writing.

What does a custom legal or healthcare SaaS platform include?

The same modules come up in almost every build. Each one below names a platform we shipped it in.

Case platform · client portal · demo data
Client portal of a case management platform: case status timeline, uploaded documents, encrypted messages with the broker
  1. Accounts, roles and portals

    Separate sign-in and scoped screens for each kind of user. The case platform runs three: client, broker and admin.

  2. Multi-tenant organisations

    Every customer company gets its own isolated data, members and billing. Built with Clerk organisations on a crew-scheduling SaaS.

  3. Subscription billing

    Stripe plans, trials, metering and invoices. The AI learning platform sells access this way.

  4. Document vault and search

    Encrypted storage in your own account, and search that answers in under 100 milliseconds across more than a million files on the legal search platform.

  5. AI drafting with citations

    A drafting pipeline that runs your instructions on the server, cites the source document for every claim, and puts a review step in front of anything a client sees. The medical chronology engine is this shape.

  6. Integrations

    Clio, MyCase, Filevine, Lawmatics, an EHR. We publish open-source connectors for the legal ones, so the sync code is already written and tested.

How do you keep client documents and AI drafting instructions confidential?

The question a lawyer or a clinician asks before uploading the first real file. The answer is the shape of the system, so here is the shape.

Request path: the browser sends a request to your API, the API reads the instruction set from a server-side secret store and the document from your encrypted bucket, calls the model provider, writes an audit row, and returns only the finished draft to the browser Browsersees the finished draft only Your APINext.js, in your account Model providertraining off, retention set to zero Instruction setserver-side secret store Documentsencrypted bucket, your cloud Audit logwho, what, when, append-only

Documents stay in your account

A bucket in your own cloud, encrypted at rest and in transit, isolated per tenant, with a retention rule you set and deletion that actually deletes.

Your instructions never reach the browser

The drafting rules that make your product yours live in a server-side secret store, versioned. A user cannot pull them out of the page, and we hold no copy after handover.

Model calls on the right plan

Through your backend, on a provider plan with training switched off and retention set to zero, and a business associate agreement where health data is involved. What that means with Claude.

Every read is logged

Who opened which document, which draft was produced from it and when. An append-only log you can hand to a reviewer, a bar or a client.

How does subscription billing work in a platform I own?

Stripe handles the money and the tax. The platform handles who gets access to what. Both are set up in your name, so the revenue never passes through us.

  • Plans, trials and upgrades, per seat or by usage.
  • Invoices, receipts and failed-payment retries that do not need you.
  • Sales tax through Stripe Tax, in the states and countries you sell in.
  • A billing page inside the product, so customers change plans without emailing you.

The AI learning platform sells access this way to clinicians who need their credits tracked.

What stack do you build on, and can my own developer take it over?

Standard tools a developer you hire next year already knows. Nothing exotic, nothing we have to be around to run.

LanguageTypeScript end to endOne language for the API, the UI and the jobs. Easier to hire for, easier to hand over.
AppNext.js on ReactServer-rendered screens, route handlers for the API, one deploy.
ServicesNode where a separate service is neededLong-running jobs, document processing, integrations.
DataPostgres with PrismaTyped schema, migrations in the repo, row-level isolation per tenant.
AuthClerkSign-in, organisations, roles, SSO when a client asks for it.
BillingStripeSubscriptions, metering, tax, invoices. The account is in your name.
HostingVercel or AWSYour account, your region, your bill.
SearchElasticsearch or Postgres full-textChosen by document volume, not by habit.

What your CTO receives

  • The repository in your GitHub organisation, with the full history.
  • A README and a runbook: how to deploy, roll back, rotate a secret, restore a backup.
  • An inventory of every environment, account and secret, and who holds access.
  • Architecture notes with the reasons behind the choices.
  • A recorded walkthrough and a handover call with the engineers who built it.

Platforms we shipped in 2024 and 2025 run on Laravel and Vue. We maintain and extend them, and if yours is one of them it stays on it. Our production checklist for those.

How does a fixed-scope SaaS build run, week by week?

The same shape for every build. Dates and acceptance criteria are written before work starts, and the people on the scoping call are the people at launch.

  1. Week 0

    Scoping call, written scope

    Your method, your users, the modules that make version one. You get the scope and the date in writing.

    Keep the scope either way
  2. Weeks 1 to 2

    Accounts, roles, admin, pipeline

    Repositories and cloud accounts in your name. Sign-in, roles, the admin screen and a deploy pipeline before any feature.

    Yours from the first milestone
  3. Milestones

    Modules, one at a time

    Each milestone has written acceptance criteria and a weekly demo on real-looking data. Same engineers from scoping to launch.

    A milestone you do not accept is not billed
  4. Launch week

    Go live, handoff

    Production checks, monitoring, the runbook and a recorded walkthrough for whoever runs it next.

    Documented for your next hire
  5. After launch

    A plan you can cancel

    A monthly plan for monitoring, patches and the next features. Cancel it and the software stays yours.

    No per-seat fee, ever

Central European hours, which overlap US Eastern mornings. Daily monitoring on your cloud, an architecture that keeps working when a third-party API is down, and a named incident contact in the agreement for the cases where something is actually down.

Two colleagues reviewing a prototype on one laptop, one pointing at the screen with a pen

I built a prototype with Lovable or Claude. Can you make it real?

Usually, yes, and the audit comes first. We read the code, run what tests exist, and check how it handles sign-in, documents and payments. What passes stays. What does not gets rewritten, with the reasons in writing. You decide, and the audit is credited to the build.

We use Claude Code on every build too. What you pay us for is the part the tool does not do: the scope, tenancy and billing design, a security review before a client document goes in, a date someone is accountable for, and a person on call after launch.

Prototype already holding health data? That has its own page.

What have you shipped for legal and healthcare companies?

Client names stay masked. The numbers and the screens are real.

[ WHO YOU'LL ACTUALLY TALK TO ]

The person on the discovery call is the person who scopes and answers for the build. No handoff to a sales team, because there isn't one.

[ FAQ ]

Questions founders ask before hiring a SaaS development company

TypeScript end to end: Next.js on React, Postgres with Prisma, Clerk for sign-in and organisations, Stripe for billing, hosted on Vercel or AWS in your own account. Every platform we started this year runs on it. Platforms we shipped in 2024 and 2025 run on Laravel and Vue, and we still maintain and extend those.

Only the parts that fail review. We start with a code audit: what passes tests and a security read stays, what does not gets rewritten, and you get the reasons in writing before anything changes. Most agencies push a full rewrite because it is easier for them. You should not pay twice for code that already works.

The engineers who scope the build are the engineers who ship it. You meet them on the scoping call, they run the weekly demo, and they hand over at launch. One of the two founders is on every build and answers for it.

The builder above gives the shape: a first version with accounts, roles, billing and one or two core modules lands in a few weeks, and a platform with documents, AI drafting and search takes longer. Anything past about eleven weeks we split into a phase one that goes live and later phases scoped on their own, so you can fund one, see it work, and decide.

Your drafting instructions live on the server, in a secrets store in your own account, and the browser only ever receives the finished output. Documents sit in a bucket in your cloud account, encrypted and isolated per tenant, with a retention rule you set. Model calls go through your backend to a provider plan with training switched off and, where health data is involved, a business associate agreement. Every read and every draft is logged with a name and a time.

Yes. Plans and trials, per-seat or usage metering, invoices and failed-payment retries, tax through Stripe Tax, and a billing page your customers can use without emailing you. The Stripe account is opened in your name on day one.

Launch week ends with a runbook, a recorded walkthrough and a handover call. After that there is a monthly plan for monitoring, patches and the next features. You can cancel it and keep the software, because the repositories, the accounts and the data were yours from the first milestone.

Central European time, which overlaps US Eastern mornings. We do not sell round-the-clock coverage we cannot staff. What you get instead: daily monitoring on your cloud so we see problems before users do, an architecture that keeps working when a third-party API is down, and a named incident contact in the agreement for the cases where something is actually down.

We use it on every build too. What you pay us for is the part the tool does not do: the scope, tenancy and billing design, a security review, a date someone is accountable for, and a person on call after launch. The tool makes a good team faster. It does not make a half-finished product safe to put client documents into.

Yes, all three, from the first paid milestone. The GitHub organisation, the Vercel or AWS account, the Stripe account, the Clerk workspace and the domain are created in your name. IP is assigned per milestone in the contract. If you ever want to stop working with us, nothing has to be moved.

Thirty minutes. Your method, your users, the weeks it takes.

You leave with a written scope and a straight answer on whether we are the right team for it. If we are not, we say so and point you somewhere better.

Scope my platform with a founder

Two builds at a time. Starts are scheduled, not sold.

Check if we're a fit