Facio Blog
Payload-powered product notes, security writing, HITL patterns, and operational guidance from the Facio runtime: long sessions, Placet approvals, audit trails, memory, providers, channels, tools, and Docker-first operations.
Showing 211-215 of 235 articles.
Product
Most AI agents start every session from scratch. Facio's layered memory architecture — combining passive context, semantic recall, inline learning, and periodic Reflection — gives agents persistent knowledge that spans sessions, tasks, and weeks of operation. Here's how it works under the hood.
Human-in-the-loop
A production database wiped. $840K in agent errors. 23% critical failure rate. The real cost of letting AI agents operate without human oversight — and why "just trust the model" is the most expensive strategy.
Engineering
ArcadeDB ships an MCP server built directly into the database engine — no external wrappers. 896 stars, 5 tools, multi-model queries (SQL, Cypher, Gremlin, MongoDB), granular security, and 70+ graph algorithms. Created by OrientDB's founder.
Security
August 2, 2026 — the EU AI Act flips from statute to enforcement. Fines up to €35M or 7% of global revenue. Here's what AI agent deployers must have in place before the deadline.
Product
Facio's built-in cron scheduler lets AI agents run recurring tasks autonomously — from daily compliance checks to hourly monitoring — without external infrastructure. Here's how the agent loop and scheduler work together, and why built-in scheduling is a hard requirement for production agent deployments.