Member-only story
Enterprise AI Gateways: Taxonomy, Capabilities, and Key Considerations
What an AI gateway does, who sells one, and why agent authorization is the unsolved problem. A 2026 field guide to LLM gateway and MCP gateway choices.
TL;DR: An enterprise AI gateway is the inline control point between an organization’s applications or agents and its AI models and tools, where every request gets authenticated, routed, metered, inspected for risk, and logged before it reaches a model or executes a tool call. These gateways have moved from thin proxies that held API keys to the enforcement point where model, inference, and agent tool traffic gets governed, and the market now sorts into seven vendor tribes that are better read as a three-axis grid of traffic plane, control function, and delivery archetype. Between 2024 and September 2026 the standalone AI security layer was almost entirely acquired into cybersecurity platforms (Cisco took Robust Intelligence, Palo Alto Networks took Protect AI for $634.5 million GAAP and then Portkey for about $140 million, F5 took CalypsoAI, SentinelOne took Prompt Security, Check Point took Lakera), and the circulating prices for most of those deals are wrong when checked against filings. Envoy AI Gateway, which reached 1.0 on 23 June 2026 with Bloomberg as co-developer and…
