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 21-25 of 236 articles.

Engineering

MCP Spotlight: Docker MCP Server — The Container Operations Bridge With `mcp://` Catalog Protocol, MCP Toolkit, and the Container-Default Reference for Agents

Jul 19, 2026Engineering

MCP Spotlight: Docker MCP Server — The Container Operations Bridge With `mcp://` Catalog Protocol, MCP Toolkit, and the Container-Default Reference for Agents

Docker's official MCP infrastructure — the Docker Hub MCP Catalog (300+ verified servers with signatures + provenance), the MCP Toolkit (isolated runtime execution via Docker MCP Gateway), and the mcp:// URL protocol for declarative installation. Apply container-ecosystem primitives (signing, isolation, secrets, orchestration) to MCP.

Product

Facio's Disaster Recovery Discipline: How AI Agent Systems Survive When the Database, the Region, or the Provider Goes Down

Jul 19, 2026Product

Facio's Disaster Recovery Discipline: How AI Agent Systems Survive When the Database, the Region, or the Provider Goes Down

Production AI agents depend on many systems: databases, model APIs, queues, message brokers, authentication services, secret stores. Any of these can fail. A database corruption takes down session state. A cloud provider incident takes down a region. A model API outage takes down reasoning. Without disaster recovery discipline, these failures cascade into extended outages. Facio's DR discipline gives AI agent systems the structural patterns to survive disasters: backups, replication, runbooks, drills.

Human-in-the-loop

HITL and the Confidence Mismatch: Why the Reviewer's Calibration Rarely Matches the Agent's

Jul 19, 2026Human-in-the-loop

HITL and the Confidence Mismatch: Why the Reviewer's Calibration Rarely Matches the Agent's

Every HITL system has two confidence scores — the agent's and the reviewer's. The scores are supposed to align. They don't. The agent is over-confident on actions that need review and under-confident on actions that don't. The reviewer is over-confident on actions they should reject and under-confident on actions they should approve. The mismatch is the calibration failure nobody measures. Here is why it happens, what it costs, and how to design for alignment.

Engineering

MCP Spotlight: Resend MCP Server — The Email Bridge With 10 Tool Groups, Send/Receive Symmetry, and the Transactional-Default Reference for Agents

Jul 18, 2026Engineering

MCP Spotlight: Resend MCP Server — The Email Bridge With 10 Tool Groups, Send/Receive Symmetry, and the Transactional-Default Reference for Agents

The official Resend MCP Server by Resend — ~25 tools across 10 domain groups covering emails, contacts, broadcasts, domains, webhooks, segments, topics, contact properties, API keys, received emails. Send/receive symmetry, Topics for GDPR preference management, hosted at mcp.resend.dev or self-hosted via NPX. MIT-licensed.

Product

Facio's Multi-Region Deployment Discipline: How AI Agents Stay Close to Users Without Losing Coordination

Jul 18, 2026Product

Facio's Multi-Region Deployment Discipline: How AI Agents Stay Close to Users Without Losing Coordination

Users are everywhere. AI agents should be close to them — for latency, for data residency, for resilience. But distributed multi-region deployments introduce consistency, coordination, and failure-mode challenges that single-region deployments never had. Facio's multi-region deployment discipline gives AI agent systems the structural patterns to operate across regions: regional autonomy for latency, asynchronous coordination for consistency, and graceful degradation for failure.