The tools behind the evidence

Project & knowledge system

Atlas

The system our project work lives in: requirements, decisions, bugs, meetings, and test runs are individual, interlinked records — a graph instead of scattered documents. Binding guardrails (constraints) apply to humans and AI agents alike: whatever violates an active rule is blocked and logged.

For client projects this means: every question about the why of a change is a query, not archaeology — including a coverage view of which requirement was last verified by which test with which result.

Agent control tower

Tower

The control center for AI agents in the development process: start, monitor, and evaluate sessions, set budgets, collect results with evidence. Tower strictly separates trusted control from untrusted content coming out of running sessions — agent work stays controllable instead of becoming a black box.

Auditable record engine

Chronik

The data foundation underneath: a record engine whose only on-disk structure is an immutable, hash-chained journal. Tampering is detected on open, every historical state can be reconstructed exactly (“what did the data look like on inspection day?”), and GDPR deletion works through key destruction without breaking the integrity chain — properties secured by crash and property-based tests, not merely claimed.