THE PERMISSION [LAYER]

Landscape

The AI-Agent Governance Vendor Map

Who sells what in agent identity, permissions, MCP security, and spend governance. The list is verified and maintained: each entry was checked against a published source, and each ownership change carries the month it closed and a link to the source we used.

76 vendors in 14 categories. Last updated November 10, 2026.

Product by product, with what each one controls and a link to the vendor: the tools directory.

The briefing

Follow changes to this map

Funding, launches and acquisitions that change this map are reported in Landscape Watch, a section of the briefing, each with an evidence link. At least once a week, free.

You confirm by email. Unsubscribe in one click.

Agent identity & NHI 9

Identity, authentication, and lifecycle for AI agents and non-human identities. Read the explainer

MCP security & gateways 7

Gateways, allowlisting, and security tooling for Model Context Protocol deployments. Read the explainer

  • Runlayer
  • Helmet Security
  • Operant AI
  • Manufact
  • Kong
  • MintMCP
  • Composio

AI security platforms 6

Broader platforms securing AI applications, models, and agents.

Agentic SOC 4

AI agents doing security operations work.

  • Torq
  • 7AI
  • Dropzone AI
  • Qevlar AI

Offensive & testing 3

Automated pentesting and adversarial testing of AI systems.

  • XBOW
  • RunSybil
  • Armadin

Exposure management 3

Discovering and managing AI-related attack surface.

  • Astelia
  • Eclypsium
  • Surf AI

LLM gateways & spend 6

Model routing, observability, and cost governance. Read the explainer

Agent observability 5

Tracing, logging, and evaluation: the audit trail for what agents actually did.

Data flow & AI-DLP 7

Controlling what data reaches agents, and what agents can leak back out.

Shadow-AI discovery 5

Finding and governing the AI tools employees adopted without asking. Read the explainer

  • Nudge Security
  • Reco
  • Grip Security
  • Island
  • Pluto Security

AI governance & compliance 6

Policy, risk, and regulatory compliance for AI programs (EU AI Act, NIST RMF). Read the explainer

Agent sandboxing (watchlist) 2

Runtime isolation for the code agents write and run. The escapes are real: in July 2026 researchers disclosed sandbox escapes in four widely used coding agents, and a CVE published in September let a sandboxed agent switch off its own confinement with a single shell command.

  • E2B
  • Daytona

Agent payments (watchlist) 3

Wallets, virtual cards, mandates, and spend limits for agents that pay for things. The rails are live: Visa ran agent purchases at independent merchants with more than 30 European issuers in July 2026, Santander and Mastercard completed Europe's first end-to-end agent payment on live banking rails in March, and an August 2026 security analysis of Google's AP2 catalogued 48 threats.

Platform giants 10

Hyperscalers and incumbents shipping agent-governance features.

Vendor-neutral: inclusion is editorial, never paid. Ownership notes link to the source we checked. Corrections: editor@thepermissionlayer.com

The Permission Layer

Get the briefing by email

At least one email a week on AI agent identity, permissions, security and spend, written for the people who approve agent access. Five-minute read, free.

You confirm by email. Unsubscribe in one click.