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 11-15 of 81 articles.
Human-in-the-loop
HITL is not always the right answer. For some workflows, it adds cost without value — latency without safety improvement. Here's a decision framework for identifying when human oversight should be removed, reduced, or replaced with automated guardrails.
Engineering
Brave's official MCP server gives AI agents 6 tools for web, news, image, video, and local search — plus AI-powered summarization. 2,000 free queries/month, rich filtering, and STDIO transport from a first-party maintainer.
Product
AI agents that can read, write, and edit files are common. Agents that understand where files belong — following project conventions, keeping roots clean, and organizing deliverables — are rare. Facio's workspace system combines a full file tool surface (read_file, write_file, edit_file, grep, glob, patch_file) with WORKSPACE.md layout rules that give agents structural awareness. Here's how it turns file access into file discipline.
Security
A single operator just ran 674 adversarial attacks against Llama Scout in three hours with an 85% success rate — using an AI red teaming agent. Traditional pentesting and manual red teams can't match this velocity. Here's what the shift to agent-orchestrated adversarial testing means for enterprise security programs.
Human-in-the-loop
When you go from one agent to ten, HITL stops being a per-agent concern and becomes a coordination problem. How do you route approvals from multiple agents to the right humans? How do you prevent approval collisions? How do you maintain a unified audit trail across a distributed agent system?