Media Shell
AI for speed. Humans for quality.
- fast delivery
- experienced software team
- AI-accelerated processes
- clear approvals & prices
- built for websites, shops & portals
- human-reviewed and owned
Senior expertise meets AI.
Digital products reach market faster.
4 experts with over 15 years of experience in software development and online marketing build websites, shops, portals and AI workflows — fast, polished and at clear prices.
-
Idea
-
AI Workflow
-
Expert Review
-
Live
Not a builder. Not a black box. AI + experience instead.
15+ years of experience
AI-accelerated
Human-checked
GDPR-compliant
Why clients pick Media Shell over builders, pure AI or classic agencies
A clear comparison: where other approaches hit limits, Media Shell combines speed, experience and clear processes.
-
15+ years software development real project experience
-
AI-automated processes less waiting time
-
Experienced humans review clean delivery
-
Clear approvals you stay in control
Clear pricing instead of agency chaos
First pick what should be built or taken over. Onboarding is the one-time start. AI Website Care on Demand is your flexible token balance afterwards. Your selection is non-binding: we personally check scope, price, timing and open points before anything starts.
1. Select package
2. Scope is reviewed
3. You receive price, timing and open points
4. Start only after your approval
1. One-time onboarding
AI Project Onboarding
Package prices: Websites
2. Token balance
AI Website Care on Demand
One-time purchase
No expiry
All onboarding packages cover 1 project. Services are offered exclusively to commercial customers. All prices are in euro and exclude statutory VAT.
Extra scope? The AI shows token usage before you approve.
Included
Project setup, technical implementation, responsive display, basic SEO, handover and the functions listed in the package.
Not automatically included
Photo/video production, legal review, third-party licenses, hosting, domains, large data migrations or special functions outside the package scope.
Additional effort
More pages, products, integrations or special requests are calculated transparently before the start. No additional costs arise without approval.
Typical combinations
Company websiteWebsite Pro + Care S for up to 6 pages, CMS, form, SEO basics and ongoing small changes.
Shop startShop Pro + Care M for products, checkout, tracking and regular care.
Existing systemExisting systems Basic/Pro + Care S or M for takeover, checks and planned development.
Team
Four specialists. One team. Fully automated.
Strategy, development, platform and customer care — every project is backed by an automated process pipeline that turns your request into results in the shortest time possible.
Founder & Strategy
Jan
Over 20 years of experience in display, affiliate marketing and business development. Combines strategic thinking with operational excellence and builds growth approaches that don't just fill reports — they deliver results.
AI & Product Engineering
Felix
Former CEO & CTO of a software company (120 employees) sold in 2022. Long-standing experience building digital products for startups and leading enterprise clients — AI-first today.
Full Stack & Platform
Martin
Builds scalable platforms, backend systems and automations for digital products. Technical focus from Django/Wagtail to API and integration engineering — with an eye on performance, maintainability and clean system architecture.
Sales & Support
Yevgen
Over 15 years of leadership experience in customer care, communication and operations. Owns sales and support — and makes sure processes are clear, response times short and client relationships reliable.
Technology stack
We take over and evolve almost any existing stack.
Whether it is an established website, shop, portal or PWA: we step into existing systems, stabilize them and build new features on the technology you already run - from CMS and commerce to backend, AI and cloud.
CMS
-
WordPress
-
Wagtail
-
Strapi
-
Drupal
-
Webflow
-
Contentful
Shop systems
-
Shopify
-
WooCommerce
-
Shopware
-
Magento
-
BigCommerce
-
Medusa
Frontend
-
React
-
Vue.js
-
Angular
-
Svelte
-
Astro
-
Tailwind CSS
Frameworks
-
Next.js
-
Nuxt
-
Remix
-
Django
-
Laravel
-
FastAPI
Backend
-
Node.js
-
Python
-
PHP
-
Go
-
Java
-
.NET
Mobile & PWA
-
React Native
-
Flutter
-
Swift
-
Kotlin
-
Ionic
-
Capacitor
APIs & integrations
-
OpenAPI
-
GraphQL
-
gRPC
-
Stripe
-
Twilio
-
Firebase
Databases
-
PostgreSQL
-
MySQL
-
MongoDB
-
Redis
-
Elasticsearch
-
MariaDB
AI & LLMs
-
OpenAI
-
Anthropic
-
Google Gemini
-
Meta Llama
-
Mistral AI
-
Hugging Face
Cloud & DevOps
-
Docker
-
Kubernetes
-
GitHub Actions
-
AWS
-
Google Cloud
-
Cloudflare
Common questions
A feature is any functional addition — from new forms and integrations through to complex logic. The AI suggests whether something is a bugfix, support request or feature, and you approve before implementation.
Corrections are small changes to your existing project. Examples include updating text, swapping images, adjusting colors, renaming buttons, changing phone numbers or refreshing a small section.
Support tickets are questions, bug checks or technical reviews. For example: why a form is not working, why a page feels slow or which setting should be changed.
Dev credits are our internal effort unit for new features or larger technical changes. 1 dev credit starts at 10 tokens. You always see token usage before approval.
Tokens are your flexible on-demand balance. 1 correction starts at 5 tokens, 1 support ticket at 50 tokens and 1 dev credit at 10 tokens. You decide per request how to use your balance.
The bot shows how many tokens are missing before approval and suggests the smallest matching token package. After successful payment, the request is approved automatically.
No. The selection is a non-binding orientation. We personally check scope, price, timing and open points, and only start once you approve the offer.
Onboarding package prices apply to the described starter scope. Token packages are a balance for care and ongoing development. Larger special features or complex risks are reviewed transparently upfront.
No. Websites and landing pages get a custom design aligned with your brand, content and goal. We use proven technical building blocks, but not a generic builder template.
We can prepare copy and image ideas with AI. Facts, approvals, brand assets, licenses and legal content such as imprint and privacy policy come from you or your legal provider.
Included is a technical baseline such as consent integration, clean forms and privacy-friendly configuration. Legal review or individually drafted legal texts are not automatically included.
A PWA runs in the browser and can be installable, but it is not a native store app. Store fees, developer accounts, final store approvals and sensitive portal workflows are scoped separately upfront.
No. Purchased tokens do not expire and are not a cash payout. They stay usable for your ongoing care and development.
No. You choose your own hosting provider, and we help with selection, setup and technical connection at the host of your choice. Domain, hosting and third-party fees are paid directly to the provider.
Yes. After a one-time onboarding we take over maintenance, support and ongoing development of your existing platform — independent of the original stack.
Small changes typically within 24-48 hours, larger features depending on scope. Complexity and approval are confirmed transparently in chat before we start.
Your project evolves one message at a time from here.
Start without obligation and see how AI Website Care on Demand works at Media Shell.
AI Update Ticker
What is moving in AI tools right now
We keep a close watch on changelogs, release notes, and AI labs so our clients can use new possibilities early.
Upcoming
Machine Learning for Robotics Software in Leipzig
-
Claude Duet handles tasks in the backgroundNEW
Anthropic introduces Claude Duet, a personal AI agent powered by Opus 5.5 that handles tasks on its own cloud computer even when users' devices are turned off.
-
Uber rolls out ChatGPT Work to over 30,000 employeesNEW
Uber is expanding ChatGPT Work from a pilot project to its entire workforce, enabling multi-step AI workflows for research, presentations, and data analysis across connected applications.
-
OCRia-1 improves document analysisNEW
LlamaIndex introduces OCRia-1, a vision-language model with four billion parameters that captures complex tables, charts, and formatting more accurately in LlamaParse.
-
ForTy identifies forest types with AINEW
Google is releasing ForTy, an AI-powered global map that uses AlphaEarth embeddings to distinguish six forest and land-use types at a resolution of ten meters.
-
Mercury 3 delivers over 1,000 tokens per secondNEW
Inception's new diffusion language model Mercury 3 generates 1,068 output tokens per second according to the company and is designed to speed up interactive AI applications such as voice assistants, learning companions, and browser agents.
-
ElevenLabs launches real-time speech translationNEW
Eleven Realtime Translation translates conversations with automatic language detection in over 90 languages and is available to enterprise customers as an alpha through ElevenAPI.
-
ScholarEvolve improves AI agents using research findingsNEW
ScholarEvolve translates scientific findings into improvements to agent software without retraining the underlying language model.
-
Mammoth-1 and Elephant-1 available on AI GatewayNEW
Vercel AI Gateway offers Baseten's Mammoth-1 and Elephant-1 decision models for text classification, request routing, and content evaluation, with free access through October 27.
-
Anthropic launches proactive AI assistant Claude DuetNEW
Claude Duet checks connected applications, identifies upcoming work and completes tasks in the background, with consequential actions requiring approval.
-
Thousands of AI agents analyze biological dataNEW
OpenAI, PerturbAI, and Scripps Research are deploying thousands of GPT-Rosalind agents to analyze data from millions of mouse brain cells in 32 minutes and generate new research hypotheses.
-
New Codex skill checks for vulnerable PHP versionsNEW
Following the AI-assisted discovery of a critical PHP vulnerability, OpenAI is releasing a Codex skill that identifies potentially vulnerable PHP versions and helps users select the necessary updates.
-
OpenAI models and Codex are coming to Google CloudNEW
OpenAI is bringing its leading models and agentic Codex workflows to Google Cloud’s Model Garden, with general availability planned for the coming weeks.
-
Mistral and NVIDIA develop AI for engineeringNEW
Mistral and NVIDIA plan to combine AI models with physics simulation and 3D generation and develop open benchmarks for engineering AI agents.
-
Cohere Rerank 5 filters relevant text passagesNEW
Cohere releases Rerank 5 in Pro and Fast variants with Context Focus, which selects relevant text passages within documents to reduce the context requirements of AI applications.
-
Eleven Realtime introduces two models for voice agentsNEW
ElevenLabs releases the Realtime v1 Mini and Max speech-to-speech models for multilingual voice agents, accessible through ElevenAgents and a WebSocket API compatible with OpenAI's Realtime specification.
-
DeepSeek V4.1 Flash joins Token PassNEW
Vercel integrates DeepSeek V4.1 Flash with the 1017 model weights into Token Pass and makes the model available through AI Gateway for coding agents and custom applications.
-
Vercel releases WebSockets for real-time AI applicationsNEW
WebSocket support in Vercel Functions is now generally available and supports AI chats, streaming, and AI Gateway's real-time models.
-
OpenAI adds twelve-hour prompt cachingNEW
The Responses API can now cache shared instructions, tools, and context for twelve hours.
-
Voxtral 3 introduces an SDK for voice agentsNEW
Voxtral 3 combines new transcription and speech synthesis models with an SDK that connects speech understanding, reasoning, and tool calls in real time.
-
AMIE collects patient information before medical appointmentsNEW
Google reports on a clinical study in which its AI research agent AMIE collected medical histories before appointments and summarized them for treating physicians.
