AI prototyping for startups
from idea to working MVP — in weeks, not months.

Discovery as fixed price from € 1,500 net, subsequent build sprints also as fixed price after discovery, optional accompanying retainer from € 1,700/month net.

Standard flow:

• First call free, 30–45 min.
• Discovery fixed price from € 1,500 net — effort and concrete price depend on complexity and stakeholder count
• Build sprints as fixed price after discovery, depending on sprint length (1, 2 or 4 weeks) and functional depth
• Accompanying retainer from € 1,700/month net with a clearly defined monthly hour quota — optional, cancellable any time

Why fixed price instead of an hour bucket: founders should know on day 1 what discovery and MVP cost — and at sprint end whether the prototype runs. Delays we carry, not you.

Validation prototype 1 week, MVP for real users 4–8 weeks — depending on complexity.

Realistic corridors:

• Validation prototype (UI-first, clickable, without real persistence) — 1 week after discovery end. Goal: collect stakeholder or investor feedback before we build a real data model
• MVP with real data model, auth, core flows — 4–8 weeks. Typically enough for the first 50–500 users
• Production build with tests, monitoring, payment, custom domain — 8–16 weeks from discovery end. Typically enough for pilot phase with pilot customers or first hundreds of paying users
• More complex use cases (marketplace, B2B multi-tenant, ML workloads) need more sprint packages — the exact number comes out of discovery

Important: "Fast" emerges in discovery, not in the sprint. A clean discovery of 1–3 weeks often shortens the build phase by weeks because stack, data-model and scope discussions are already done.

We combine Cursor, Claude Code, Lovable, v0 and Bolt with classical senior engineering — not "one of them for everything".

When which tool:

• Cursor — primary IDE for senior engineering, larger codebases, refactorings, test generation
• Claude Code — agentic coding sessions, autonomous feature branches, migration paths, test suites
• Lovable / v0 — UI-first validation prototypes when fast visual feedback against stakeholders or investors is needed
• Bolt — rapid stack bootstraps when frontend, backend and DB should be set up together in one session
• Classical engineering (TypeScript, manual architecture, code review, Git discipline) — foundation for anything that should become production-ready

What we consciously don’t do: "vibe coding" without an architecture plan. If a Lovable prototype is to go productive, a refactoring phase belongs to it that sets up the data model, auth logic and hosting architecture cleanly. Otherwise the startup pays for the later rebuild.

Yes — you retain 100 % of the code, data and hosting accounts at every point in time.

Concretely:

• Repository lives in your GitHub/GitLab account (or we set it up there)
• Cloud accounts (Fly.io, Hetzner, Cloudflare, Stripe, Resend) run in your name — we work as invited team members, never via agency-owned wrappers
• Domain & DNS run in your name
• License: full source-code transfer in the work agreement — no sub-licensing, no lock-in
• Sub-processor list as part of the handover documentation so you know which third parties process data

When the collaboration ends, you keep everything and can continue yourself at any time or use a different partner — no "you can only continue with us because otherwise the code won’t run any more".

Default: Next.js + TypeScript + PostgreSQL + Fly.io or Hetzner + Stripe + Resend.

Why this stack:

• Next.js (App Router) — one codebase for SEO-capable frontend, API routes, server components, built-in image optimisation. Ideal for MVPs that scale later
• TypeScript — same language from frontend to API, low refactor pain, excellent AI coding support in Cursor and Claude Code
• PostgreSQL — relational, ACID, mature, runs everywhere (Neon, Supabase, Hetzner, Fly.io)
• Fly.io (region FRA) for full container control and EU hosting; Hetzner (Falkenstein/Nuremberg) for cost-efficient self-hosted setups
• Stripe — standard payment stack, best DX, EU-capable
• Resend — modern transactional email with good deliverability and EU data residency

When we deviate:

• Mobile-first use case → Expo / React Native
• Heavy ML or CV → Python backend (FastAPI) with Next.js frontend
• B2B SaaS with enterprise SSO from day 1 → early switch to Clerk or WorkOS for auth
• Existing codebase in Vue, Astro, Laravel with active senior team → we build inside it, instead of flipping

What we don’t use: Wix, Webflow or comparable no-code platforms for productive web apps — they scale neither operationally nor under GDPR requirements.

Default: EU hosting via Fly.io region Frankfurt or Hetzner Cloud (Falkenstein/Nuremberg) — both GDPR-compliant and with clear sub-processor documentation.

Hosting options we actively use:

• Fly.io (Frankfurt FRA) — container stack with multi-region capability, built-in Postgres and volume management; ideal for Next.js apps with dynamic workloads
• Hetzner Cloud (DE-Falkenstein, DE-Nuremberg, AT-Vienna) — cost-efficient VPS and managed Postgres setups; good for self-hosted architectures
• Cloudflare Pages — when static marketing sites without sensitive data are built; with awareness that edge functions may route via US
• Raidboxes — for WordPress-based marketing sites alongside the product stack

What we deliver technically / strategically on the GDPR/FADP question:

• Hosting architecture in EU regions
• Data-flow classification for data-processing agreements
• Tracking setup with Consent Mode v2 (default GA4 / GTM, alternatively Plausible for analytics only)
• Sub-processor overview for Stripe, Resend, Sentry, OpenAI/Anthropic APIs

What we don’t provide: legally binding GDPR or FADP assessment, privacy policies, cookie-banner texts, DPIAs, communication with authorities. That belongs in professionally appropriate hands — we actively recommend the parallel involvement of your lawyers or data-protection officers.

Three models, depending on your setup:

• In-house take-over — we hand over repository, architecture documentation, onboarding walkthrough for the first in-house engineer and remain available for 4–8 weeks for targeted questions via Slack Connect. Ideal if you already have a CTO or senior engineer on the team
• Accompanying retainer from € 1,700/month net — we stay on board for bugfixes, smaller feature extensions, hosting care and performance optimisation. Clearly defined monthly hour quota, response time 24 h on business days as default. Ideal for founder teams without in-house engineering
• Sprint-on-demand — when the MVP runs and only needs development in larger bursts (e.g. Q1: payment integration, Q3: B2B extension), we work in planned fixed-price sprints without an ongoing retainer

In all three models: you retain 100 % of the code, data and hosting accounts. You can switch mode at any time or bring in a different partner — no lock-in.

Standard response to every request: within 24 hours on business days — even with no faster SLA contractually agreed.

What is possible beyond that:

• Sprint phases — daily stand-ups or async updates via Slack Connect, response usually significantly faster than 24 h
• Retainer with faster SLAs as a contractual add-on — e.g. 4 h response in CET business hours or on-call standby with under 30 minutes via SMS for critical production issues
• Emergency escalation for production outages is contractually agreed, not a "costs extra in the moment of emergency" model

What we don’t promise: immediate response outside agreed SLAs. Clear response-time contracts are more professional than suggested 24/7 availability without real backup structures.

Three structural differences:

• Operator practice — for over 15 years we have built digital products ourselves (own shops, brands, SaaS tools). When we discuss tech stack, scaling sequence or hosting architecture, it is from engagements where we have walked the same step ourselves multiple times. Pure freelancer collectives or no-code boutiques generally lack this operator depth
• Senior lead with fixed backup — you get a fixed senior lead and a second senior team member as backup. No rotating juniors, no account-manager layer, no "the one freelancer is currently on a different mission"
• End-to-end responsibility instead of piece-work — we deliver discovery, build, hosting, handover and optional accompanying retainer from one source. With freelancer collectives, in the end often nobody carries responsibility for production stability

When the freelancer or no-code path fits better: when the project is small enough that a 1-week Lovable session suffices, or when you already have a CTO and engineering team that only needs spot reinforcement. In both cases we say so openly.

Clear delineation — we are engineering and product accompaniment, not an investment, patent or marketing studio.

What we don’t do:

• Pitch deck design or investor storyboard layouts — we deliver the prototype that makes the pitch credible; deck design belongs with specialist pitch-deck studios
• Investment or funding advice — we accompany technically but assess neither term sheets nor cap tables nor valuation multiples. That belongs with specialist startup lawyers or venture consultancies
• Patent or trademark filing — belongs with patent attorneys
• Performance marketing or growth-hacking operations — we build the tracking setup but do not steer daily advertising campaigns
• Legally binding assessment — GDPR/FADP evaluation, terms drafting, competition-law review. Belongs with lawyers
• Classic market research — representative consumer surveys, brand-tracking studies. Belongs with market-research institutes

We say so actively: when a request clearly belongs in another discipline, we refer transparently — no invented engagements, no selling of disciplines we are not strong in.