For enterprises handling sensitive assets under GDPR, financial regulations, or classified protocols, Palimpsests provides a trusted, local-first platform for artificial intelligence inference. By executing models entirely on-premises, Palimpsests guarantees that no data is ever uploaded or transmitted to cloud providers. Every turn of a conversation or tool execution is stamped onto a cryptographic, hash-chained record, creating a permanent, tamper-evident audit trail that shows exactly how the AI operated without revealing confidential context to unauthorized parties.

Engineered for optimal resource efficiency on modest hardware, Palimpsests preserves computed attention states across sessions, allowing complex prompts and multi-step tool calls to execute without repetitive context pre-filling. The system relies on the open PALA-1 v1.0 logging standard, which can be verified completely offline without decryption keys or external witness servers. Fully compliant with Apache-2.0 open-source licensing and designed to meet strict logging regulations, Palimpsests bridges the gap between high-performance local AI capabilities and absolute operational transparency.