September 2026 Summaries

Barndoor’s September 2026 release introduces Barndoor Automations, formerly Diaphora, an acquired platform for building model-agnostic AI-agent workflows that operates on Barndoor’s access-control and audit infrastructure. Automations connect through MCP and OAuth, while each custom workflow, called a Blueprint, can be governed as an individual tool and can route models through the Barndoor LLM Gateway. MCP Governance enhancements include customizable policy-denial messages, wildcard and pattern-based field-level controls for resources such as Slack channels and Google Drive files, and an MCP server that lets administrators query live governance data in natural language and draft policies for review. LLM Gateway updates require users to provide a reason for changes to providers, credentials, and routes, and add daily cost breakdowns to support spending reconciliation. The release also makes dark mode available to all users and adds governance support for MCP connectors including Amperity, Calendly, Postman, PayPal, and Smartsheet.

An AI gateway is presented as a centralized governance layer for AI traffic that verifies identities, enforces access and data-protection policies, manages costs, routes requests, and records activity across model calls and agent interactions with business tools. Unlike conventional API gateways, it is designed to interpret AI-specific elements such as prompts, tokens, model selection, sensitive information, and tool-call parameters. As AI agents increasingly perform actions such as accessing email, updating CRM systems, querying databases, and filing tickets, gateways need agent-specific identities, delegated user permissions, per-tool authorization, and action-level audit logs. The text distinguishes AI gateways, which primarily govern application-to-model traffic, from MCP gateways, which control agents’ access to tools through the Model Context Protocol, while arguing that organizations using agents need coordinated controls for both. It recommends gateways when AI use expands across teams, providers, regulated data, or connected business systems, while noting limitations including bypassed traffic, availability dependencies, added latency, and the inability to eliminate unsafe model behavior or replace human oversight for high-stakes decisions.

Barndoor AI has acquired Diaphora, creator of the open-source Frags AI workflow engine, to combine governed enterprise access controls with technology for building repeatable, predictable AI automations called Blueprints. Blueprints define a series of tool and data interactions, limiting where large language models can make judgments while ensuring other steps execute consistently, stop when required actions fail, and provide visibility into activity, changes, and costs. The combined platform aims to help enterprises move beyond AI pilots by allowing authorized employees to discover and run role-approved automations without needing direct access to every underlying system or separate provisioning for each workflow. Potential uses include completing operational actions such as correcting invoices and assembling information from multiple business systems for customer reviews. Diaphora’s full team will join Barndoor, while Frags and its Frags Modeling Language will remain open source and supported by Barndoor.