Not sure which mode fits? A short system and architecture clarification makes the goal, risks, and sensible scope visible before a larger engagement begins.
More delivery capacity – without automating responsibility.
Entwicklerherz orchestrates specialized AI agents for code analysis, implementation, tests, and documentation. The agents operate within a clear technical frame; an experienced software architect holds context, decisions, and outcome together.
AI agents handle code analysis, implementation, tests, and documentation – within a clear technical frame.
You are not hiring an autonomous bot team. You work with one accountable senior engineer whose delivery capacity is extended by agents.
No autonomous bot team: you work with one accountable senior engineer.
01Understand and scopeSystem context, business rules, risks, and acceptance criteria are clarified together. This creates a dependable working frame.Senior-led
02Orchestrate agentsSpecialized agents analyze the codebase, handle bounded tasks, and produce reviewable changes with tests and documentation.AI accelerates
03Review and decideArchitecture, security, maintainability, and business effect are reviewed by a human. Critical decisions stay explicit and traceable.Senior-owned
04Integrate and shipThe change is integrated into build, operations, and existing team workflows with the evidence required for production approval.Production together
Anzeigenpilot was a complete commerce platform, not merely an AI listing generator. Sellers turned clothing photos into listings, managed inventory, and grouped items in personal closets or time-limited drops. Buyers searched items and sellers, maintained wishlists, asked questions, and completed purchases through integrated communication and payment workflows.
ArchitectureNext.js 16 · PostgreSQL/Supabase · role and permission model · OpenAI Vision · search and wishlists · messaging workflows · Stripe Connect · OAuth 2.0 with PKCE · MCP
SchemaPath combined practice management and digital schema therapy in one shared role and process model. Therapists managed patients, appointments, and groups, communicated securely, and assigned tasks, materials, exercises, and learning sets. Patients used journals, mode tracking, messaging, and interactive content between sessions; activity and progress flowed back into the therapeutic overview.
ArchitectureNext.js 16 · PostgreSQL/Supabase · role-based access · WebAuthn · calendar and group workflows · messaging and assignments · modular exercise and learning-set engine · Azure OpenAI · 7 languages
Therapendo guided healthcare practices from template selection to a published website. Content could be edited directly without technical expertise; hosting, a Therapendo subdomain or custom domain, and contact forms were integrated. Patient inquiries could be delivered to the practice with end-to-end encryption.
ArchitectureNext.js 16 · PostgreSQL/Supabase with row level security · TipTap/MDX · template and publishing engine · custom domains · X25519 + XSalsa20-Poly1305 · Stripe · 29 languages
// Entwicklerherz Lab
Do not just talk about AI engineering. Build tools for it.
The Lab creates engineering tools with a clearly stated maturity level. Changeproof is a working prototype tested against real codebases – not yet an independent product.
The Changeproof CLI analyzes a Git diff together with a ticket or specification. It creates a local Change Passport that maps requirements to changed files and tests, exposes open risks, and adds a verifiable integrity hash.
Reviewers get one artifact showing what was intended, what changed, which evidence exists, and what remains open before approval.