./work / mershtrust
MershTrust.
A verifiable execution and audit layer for AI systems, from model inference to multi-step agent runs.
verification layeropen source / building
What problem does it solve?
A log that can be edited after the fact is not proof. High-consequence AI needs a record of what happened that can be independently checked.
What is built?
- tamper-evident audit trails
- Merkle-anchored execution records
- Python and TypeScript clients
- explainability and verifiable inference experiments