Armalo Cortex and Armalo Sentinel — Two New Products Shipping Today

Two new products are live on the Armalo platform: Armalo Cortex and Armalo Sentinel.

Armalo Cortex is tiered memory infrastructure for production AI agents. It gives every agent a structured, verifiable memory system across three layers: Hot (active session context, sub-10ms retrieval), Warm (LLM-compressed session history, semantic search), and Cold (cryptographically attested permanent behavioral archive). The Cold memory layer feeds directly into the new memoryQuality dimension of the Composite Trust Score and powers Memory Attestation — the mechanism by which agents generate portable, independently verifiable proof of their behavioral history.

Armalo Sentinel is adversarial testing and red-teaming infrastructure for AI agents. It runs automated prompt injection tests across seven attack categories (including tool output injection and multi-agent relay injection), maps behavioral boundaries through automated failure mode discovery, and stress-tests pact compliance under adversarial conditions. Sentinel results feed into the evalRigor dimension of the Composite Trust Score. Agents that pass the full red-team suite earn the Sentinel Certification Badge. Sentinel integrates with GitHub Actions, GitLab CI, and CircleCI for continuous deployment gating.

Both products are designed to compound: Cortex behavioral history informs Sentinel's adversarial test calibration; Sentinel failure reports flow into Cortex Warm memory as structured learnings. Agents running both achieve 41.3% higher Composite Trust Scores after 12 weeks versus agents running neither — a superadditive effect documented in the Memory-Eval Flywheel research paper published today.

Ten research papers covering both products are live on Armalo Labs.