Services
We offer three consulting services, each with a defined scope, timeframe and pricing model. For a complete application, see our custom software offer.
We build Approval and Reynt on our own open-source platform, and we run Reynt in production. These services bring that experience to a system your team already owns. We work in German and English, remotely across the EU or on site around Stuttgart.
We plan data protection and tenant isolation from the start: row-level security in PostgreSQL separates tenants, and personal data is encrypted per tenant, with deletion through key destruction.
An independent review of your existing system.
We review your code, data model and operations, and talk to the people maintaining the system. We look at module boundaries, tenant isolation, potential failures and migrations that may become harder to complete over time.
Useful before deciding to rewrite a system, changing its vendor or handing it to a new team.
Introduce coding agents with automated checks and a clear review process.
We use AI coding agents daily to develop Reynt. With your team, we set up automated architecture checks, tests and a human review process, then work alongside you to put them into practice.
For teams that want to use coding agents while maintaining code quality.
Plan the architecture before development begins.
We help choose technologies, define service and module boundaries, and design the data, tenancy and permissions models. We document the trade-offs and add checks to keep the implementation consistent as the team grows.
For teams turning a prototype into a production system.
AI-assisted development in detail
Coding agents can produce more code than a person can review at once. These six practices help us check their work and keep changes small enough to review.
01
A static test rejects framework imports in the domain logic. When we moved all four Reynt APIs from Elysia to Hono, the changes were confined to route files; the domain code stayed the same.
02
PostgreSQL enforces tenant isolation through row-level security with FORCE on every tenant-scoped table. This provides a second layer of protection if application code omits a tenant filter.
03
We parse input at the system boundary and pass it on as typed values. This avoids repeated validation and lets the compiler catch invalid combinations before the code runs.
04
Agent-written tests can repeat mistakes from the implementation. Our integration tests run against a real PostgreSQL instance in CI, so they also check actual database behaviour.
05
Agents produce code faster than people can review it. We keep changes small and self-contained, describe their purpose in commits and record decisions in a work journal. That gives reviewers enough context to assess each change.
06
We review new dependencies before adding them. Infrastructure changes go through GitOps, secrets are stored encrypted and kept out of agent context, and release images are signed.
Questions
We would first look at what went wrong in your codebase. Automated architecture checks, relevant tests and manageable reviews can catch problems earlier. Setting those up is part of the engagement; changing the tool alone may not help.
The code is written in your repository under your contract. Before using a tool, we check its data-use terms and configure the applicable training exclusions. We document the settings and contractual terms for each tool.
We review how each tool processes data, including the processing agreement and options for EU processing. Customer data, personal data and production data are kept out of the assistant’s context.
No. Developers spend more time designing, reviewing and making decisions, and less time writing code by hand. Review capacity can become the bottleneck, so a team may need to improve its review process before it sees a productivity gain.
Reynt is a booking platform in production that we built this way. You can read the documentation of its underlying MIT-licensed libraries, whose source is public from November 2026, and our descriptions of the engineering patterns, with code examples and trade-offs.
They need different approaches. Coding agents support your development team. AI features in a product affect your users and may raise regulatory questions. Our oversight kit lets people review those model proposals and records their decisions. Legal requirements need a separate assessment by your legal adviser.
Terms
Scope
We discuss these boundaries in the first call, so both sides know what to expect.
Tell us what you want to build or improve and when you need it. A short email is enough to start. The person who would work with you usually replies within one or two working days.