AIR vs Zenity for coding agents and the skills, plugins, and MCP servers they install

AIR’s launch coverage says its whitelist filters out about 27% of the AI add-ons it evaluates. If that rate holds anywhere near your environment, every developer who installs a handful of skills or MCP servers into a coding agent is likely running something nobody vetted. Both AIR and Zenity now sell controls for that problem, and they enter it from different starting points.

AIR came out of stealth in September 2026 around the add-on supply chain. AIR Filter vets skills, plugins, MCP servers, and subagents before they are installed and after updates. AIR Control governs sanctioned and shadow agents. AIR Defend protects agent actions at runtime, and AIR Marketplace supplies pre-vetted add-ons. Zenity started with agent governance across low-code and SaaS agents and now documents coverage for Claude Code and Cowork, ChatGPT Codex, GitHub Copilot, and Cursor through native agent hooks and the Zenity MCP gateway. AIR’s capabilities are recent launch claims, stated rather than demonstrated.

Buy AIR when the most urgent risk is what developers install into agents, and you want every add-on vetted before install and drawn from a trusted catalog. Buy Zenity when you need one platform governing agents across Copilot Studio, Salesforce, ServiceNow, and cloud stacks as well as coding agents on laptops.

Overview

	AIR	Zenity
Primary use case	Securing the AI add-on supply chain for agents, with fleet governance and runtime protection around it	AI agent security and governance across SaaS, cloud, homegrown, and coding and personal agents
Day-to-day workflow	Review agents found across the fleet, vet add-ons before install and after updates, steer developers to the AIR Marketplace, and respond to runtime detections	Discover agents and the MCP servers and tools they reach, set boundaries, and investigate threat signals mapped to OWASP and MITRE ATLAS
Main buying consideration	No public pricing. Sold by demo. Free add-on and agent scanners are on the site	No public pricing on zenity.io. Sold by demo and quote

We reviewed AIR’s homepage, its September 2026 launch post, and launch coverage, plus Zenity’s homepage, the coding and personal agents page, the MCP security page, and its August 2026 funding release, on 14 Oct 2026. We did not run either product.

AIR

Zenity

Editions and pricing

Neither vendor prints a price, so the difference is scope: which agents and which layer of risk the contract covers.

	AIR	Zenity
How it is sold	AIR Control, AIR Filter, AIR Defend, and AIR Marketplace as one platform	One platform covering AI observability, AISPM, exposure management, boundaries, agentic identity, MCP security, and AI detection and response
Free tooling	Scan your add-ons and scan your agent tools on the site	Security assessment hub and Zenity Labs research
Agent coverage named	Coding agents and their skills, plugins, MCP servers, and subagents	Claude Code and Cowork, ChatGPT Codex, GitHub Copilot, Cursor, Microsoft 365 Copilot, Copilot Studio, Salesforce Agentforce, ServiceNow, Amazon Bedrock, Google Vertex AI, and more

Add-ons and MCP servers

Both put a control point between agents and the tools they load. They differ on whether the control sits at install time or at call time.

	AIR	Zenity
Control point	AIR Filter vets each add-on before deployment, after updates, and while it runs	Zenity MCP gateway sees every server, tool, and call, with nothing changed on the agent side
Policy action	Block unvetted add-ons and supply trusted ones through AIR Marketplace	Allow, modify, or block a tool call based on live context
Risks named	Hidden behavior, prompt injection, agent hijacking, excessive permissions, external instruction loading, data exfiltration, supply chain takeover	Tool poisoning, tool drift or rug pulls, shadow MCP servers, secret exfiltration, and malicious package installs

Runtime protection on endpoints

Both watch agents while they act. Coverage beyond coding agents is where the documented scope diverges.

	AIR	Zenity
Runtime layer	AIR Defend detects, responds, and protects in real time on agent actions	Native agent hooks monitor what code an agent reads, what commands it runs, what it installs, and what it sends to models
Fleet governance	AIR Control holds sanctioned and shadow agents to policy on config, identity, and permissions	Discovery of every coding and personal agent across laptops and desktops, plus agent inventory across SaaS and cloud
Non-coding agents	Not confirmed in public documentation	Agentic SaaS, cloud, and homegrown agents are documented use cases

Choose AIR when developer-installed skills, plugins, and MCP servers are the exposure you need to close first, and you want a vetted catalog plus checks before install and after every update. Choose Zenity when the same program also has to govern Microsoft, Salesforce, ServiceNow, and cloud-built agents, and an MCP gateway that rules on each tool call fits your architecture.