Facio Blog

Practical notes on human-reviewed AI agents.

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 91-95 of 235 articles.

Product

Facio's Graceful Degradation: How AI Agents Keep Working When the Stack Around Them Breaks

Jun 29, 2026Product

Facio's Graceful Degradation: How AI Agents Keep Working When the Stack Around Them Breaks

Production AI agents operate in hostile environments. MCP servers go offline, APIs return 500s, databases time out, networks drop mid-call. The agent that fails on the first hiccup is a prototype. The agent that degrades gracefully — falling back to alternatives, retrying intelligently, escalating when needed, and continuing with what it has — is a production system. Facio's graceful degradation patterns are the structural discipline that keeps agents working when the world around them isn't. Here's how the patterns work.

Human-in-the-loop

HITL Is a Contract: The Seven Clauses Between the Agent, the Reviewer, and the System

Jun 29, 2026Human-in-the-loop

HITL Is a Contract: The Seven Clauses Between the Agent, the Reviewer, and the System

HITL is sold as a feature, but it's actually a contract between an AI agent, a human reviewer, and the system that connects them. When that contract breaks, every party loses. Here's the seven-clause contract every HITL deployment needs in writing — and what happens when each clause is missing.

Engineering

MCP Spotlight: Anthropic Filesystem MCP Server — Sandboxed Read/Write, Allow-Listed Directories, and the Reference Architecture for File Access

Jun 28, 2026Engineering

MCP Spotlight: Anthropic Filesystem MCP Server — Sandboxed Read/Write, Allow-Listed Directories, and the Reference Architecture for File Access

The Anthropic Filesystem MCP Server is the reference implementation for safe file access in the MCP ecosystem — ~14 tools (read, write, edit, search, move, delete), MIT-licensed, with strict --allowed-directory sandbox + symlink resolution at the boundary. The architectural template every local-resource MCP server should copy.

Product

Facio's Context Window Discipline: How AI Agents Stay Sharp When Conversations Run for Hours

Jun 28, 2026Product

Facio's Context Window Discipline: How AI Agents Stay Sharp When Conversations Run for Hours

An AI agent's context window is its working memory — and it's finite. Long sessions accumulate context: every tool call result, every message, every retrieved fact stays in the window until the session ends. The agent gets slower, more expensive, and eventually confused as the context fills. Facio's context window discipline gives agents the structural patterns to stay sharp across hours of work: selective loading, aggressive summarization, strategic forgetting, and checkpoint-driven compaction. Here's how the discipline works.

Human-in-the-loop

HITL and Model Versioning: How Approval Patterns Change When the LLM Changes

Jun 28, 2026Human-in-the-loop

HITL and Model Versioning: How Approval Patterns Change When the LLM Changes

You upgraded the LLM from claude-opus-4-7 to the next generation. The HITL override rate spiked from 4% to 12% overnight. The same reviewer pool is rejecting 3× more actions. The same policy is firing 5× more synchronous reviews. Model upgrades don't just affect outputs — they affect the entire HITL calibration. Here's how to manage the version transition.