NEW: UNIFY AI AGENT SPEND, SECURITY & GOVERNANCE IN ONE PLACE KONG JOINS SERVICENOW CONTROL TOWER ALLIANCE GARTNER® NAMES KONG A LEADER FOR 7TH YEAR GET THE API + AI SUMMIT 2026 RECAP KONG NAMED A LEADER IN FORRESTER WAVE: API MANAGEMENT SOFTWARE
  • [Why Kong ](/company/why-kong)Why Kong
  • _API & AI CONNECTIVITY TECHNOLOGIES_
    The AI Connectivity Platform
    []
    API ManagementAI ManagementEvent ManagementMonetization
    Migration Services
    API Advisory Services + Forward Deployed EngineersNEW
    • RUNTIMES
    • [API Gateway ](/products/kong-gateway)API Gateway
    • [AI Gateway HOT](/products/kong-ai-gateway)AI Gateway HOT
    • [Event Gateway ](/products/event-gateway)Event Gateway
    • [Service Mesh ](/products/kong-mesh)Service Mesh
    • [Context Mesh ](/products/kong-konnect/features/context-mesh)Context Mesh
    • [Kong Operator ](/products/kong-operator)Kong Operator
    • [Webhook Engine ](/products/kong-webhook-engine)Webhook Engine
    • CORE SERVICES
    • [AI Registry NEW](/products/mcp-registry)AI Registry NEW
    • [Catalog ](/products/kong-konnect/features/api-service-catalog)Catalog
    • [APIOps & Automation ](/products/apiops-automation)APIOps & Automation
    • APPS & AI AGENTS
    • [Developer Portal ](/products/kong-konnect/features/developer-portal)Developer Portal
    • [Usage Billing & Metering ](/products/kong-konnect/features/usage-based-metering-and-billing)Usage Billing & Metering
    • [Observability ](/products/kong-konnect/features/api-observability)Observability
    • [KAi Agent ](/products/kong-konnect/features/kai-ai-agent)KAi Agent
    • [Identity & Access ](/products/governance-and-identity)Identity & Access
    • [AI Cost Management ](/products/ai-cost-management)AI Cost Management
    DEVELOPER TOOLS
    [Insomnia ](https://insomnia.rest/)Insomnia [Plugins ](https://developer.konghq.com/plugins/)Plugins [Volcano ](https://volcano.dev/)Volcano [Kong MCP ](https://developer.konghq.com/konnect-platform/konnect-mcp/)Kong MCP [Documentation ](https://docs.konghq.com/)Documentation [Open Source ](/community)Open Source
      • FOR PLATFORM TEAMS
      • [Developer Platform ](/solutions/building-developer-platform)Developer Platform
      • [Kubernetes and Microservices ](/solutions/build-on-kubernetes)Kubernetes and Microservices
      • [Observability ](/solutions/observability)Observability
      • [Service Mesh Connectivity ](/solutions/service-mesh-connectivity)Service Mesh Connectivity
      • [Kafka Event Streaming ](/solutions/kafka-stream-api-management)Kafka Event Streaming
      • FOR EXECUTIVES
      • [Open Banking ](/solutions/open-banking)Open Banking
      • [Legacy Migration ](/solutions/legacy-api-management-migration)Legacy Migration
      • [Platform Cost Reduction ](/solutions/api-platform-consolidation)Platform Cost Reduction
      • [Kafka Cost Optimization ](/solutions/reduce-kafka-cost)Kafka Cost Optimization
      • [API Monetization ](/solutions/api-monetization)API Monetization
      • [AI Monetization ](/solutions/ai-monetization)AI Monetization
      • FOR AI TEAMS
      • [AI-Native API Management ](/solutions/ai-native)AI-Native API Management
      • [Agent Gateway ](/agent-gateway)Agent Gateway
      • [Trigger Agents from Events ](/solutions/event-driven-ai-agents)Trigger Agents from Events
      • [AI Governance ](/solutions/ai-governance)AI Governance
      • [AI Security ](/solutions/ai-security)AI Security
      • [Token Cost Management ](/solutions/ai-cost-optimization-management)Token Cost Management
      • [Agentic Infrastructure ](/solutions/agentic-ai-workflows)Agentic Infrastructure
      • [MCP Production ](/solutions/mcp-production-and-consumption)MCP Production
      • [MCP Traffic Gateway ](/solutions/mcp-governance)MCP Traffic Gateway
      • FOR DEVELOPERS
      • [Mobile App API Development ](/solutions/mobile-application-api-development)Mobile App API Development
      • [GenAI App Development ](/solutions/power-openai-applications)GenAI App Development
      • [API Gateway for Istio ](/solutions/istio-gateway)API Gateway for Istio
      • [Decentralized Load Balancing ](/solutions/decentralized-load-balancing)Decentralized Load Balancing
      • BY INDUSTRY
      • [Financial Services ](/solutions/financial-services-industry)Financial Services
      • [Healthcare ](/solutions/healthcare)Healthcare
      • [Higher Education ](/solutions/api-platform-for-education-services)Higher Education
      • [Insurance ](/solutions/insurance)Insurance
      • [Manufacturing ](/solutions/manufacturing)Manufacturing
      • [Retail ](/solutions/retail)Retail
      • [Software & Technology ](/solutions/software-and-technology)Software & Technology
      • [Transportation ](/solutions/transportation-and-logistics)Transportation
    NEW
    Kong for Startups
    Apply for $100k credits & 50% off AI Gateway
  • [Pricing ](/pricing)Pricing
      • DOCUMENTATION
      • [Kong Konnect ](https://developer.konghq.com/konnect/)Kong Konnect
      • [Kong Gateway ](https://developer.konghq.com/gateway/)Kong Gateway
      • [Kong Mesh ](https://developer.konghq.com/mesh/)Kong Mesh
      • [Kong AI Gateway ](https://developer.konghq.com/ai-gateway/)Kong AI Gateway
      • [Kong Event Gateway ](https://developer.konghq.com/event-gateway/)Kong Event Gateway
      • [Kong Insomnia ](https://developer.konghq.com/insomnia/)Kong Insomnia
      • [Plugin Hub ](https://developer.konghq.com/plugins/)Plugin Hub
      • EXPLORE
      • [Blog ](/blog)Blog
      • [Customer Stories ](/customer-stories)Customer Stories
      • [Learning Center ](/blog/learning-center)Learning Center
      • [eBooks ](/resources/e-book)eBooks
      • [Reports ](/resources/reports)Reports
      • [Demos ](/resources/demos)Demos
      • [Videos ](/resources/videos)Videos
      • EVENTS
      • [API + AI Summit ](/events/conferences/api-ai-summit)API + AI Summit
      • [Webinars ](/events/webinars)Webinars
      • [User Calls ](/events/user-calls)User Calls
      • [Workshops ](/events/workshops)Workshops
      • [Meetups ](/events/meetups)Meetups
      • [See All Events ](/events)See All Events
      • FOR DEVELOPERS
      • [Engineering Blog ](/blog/engineering )Engineering Blog
      • [Get Started ](https://developer.konghq.com/)Get Started
      • [Community ](/community)Community
      • [Certification ](/academy/certification)Certification
      • [Training ](https://education.konghq.com)Training
      • COMPANY
      • [About Us ](/company)About Us
      • [We're Hiring! ](/company/careers)We're Hiring!
      • [Press Room ](/company/press-room)Press Room
      • [Contact Us ](/company/contact-us)Contact Us
      • [Kong Partner Program ](/partners)Kong Partner Program
      • [Enterprise Support Portal ](https://support.konghq.com/s/)Enterprise Support Portal
      • [Documentation ](https://developer.konghq.com)Documentation
  • [](/search)
  • [Login](https://cloud.konghq.com/login)Login
  • [Book Demo](/contact-sales)Book Demo
  • [Get Started](/products/kong-konnect/register)Get Started
[Blog](/blog)Blog
  • [AI Gateway ](/blog/tag/ai-gateway)AI Gateway
  • [AI Security ](/blog/tag/ai-security)AI Security
  • [AIOps ](/blog/tag/aiops)AIOps
  • [API Security ](/blog/tag/api-security)API Security
  • [API Gateway ](/blog/tag/api-gateway)API Gateway
|
    • [API Management ](/blog/tag/api-management)API Management
    • [API Development ](/blog/tag/api-development)API Development
    • [API Design ](/blog/tag/api-design)API Design
    • [Automation ](/blog/tag/automation)Automation
    • [Service Mesh ](/blog/tag/service-mesh)Service Mesh
    • [Insomnia ](/blog/tag/insomnia)Insomnia
    • [Event Gateway ](/blog/tag/event-gateway)Event Gateway
    • [View All Blogs ](/blog/page/1)View All Blogs
We're Entering the Age of AI Connectivity [Read more](/blog/news/the-age-of-ai-connectivity)Read moreProducts & Agents:
    • [Kong AI Gateway](/products/kong-ai-gateway)Kong AI Gateway
    • [Kong API Gateway](/products/kong-gateway)Kong API Gateway
    • [Kong Event Gateway](/products/event-gateway)Kong Event Gateway
    • [Kong Metering & Billing](/products/kong-konnect/features/usage-based-metering-and-billing)Kong Metering & Billing
    • [Kong Insomnia](/products/kong-insomnia)Kong Insomnia
    • [Kong Konnect](/products/kong-konnect)Kong Konnect
  • [Documentation](https://developer.konghq.com)Documentation
  • [Book Demo](/contact-sales)Book Demo
  1. Home
  2. Blog
  3. Enterprise
  4. Agent, Identity, and MCP Governance in One Control Layer
[Agentic AI](/blog/tag/agentic-ai)Agentic AI
November 17, 2026
13 min read

# Agent, Identity, and MCP Governance in One Control Layer

Kong

Somebody on your team just wired an agent to the CRM, a Slack workspace, and three Model Context Protocol (MCP) servers nobody reviewed. It works great but nobody's quite sure who owns it.

That's the state of agent adoption right now. Only 13% of organizations think they have the right AI agent governance in place, according to[ _Gartner's agent sprawl research_](https://www.gartner.com/en/newsroom/press-releases/2026-04-28-gartner-identifies-six-steps-to-manage-artificial-intelligence-agent-sprawl) _Gartner's agent sprawl research_. The same research projects that the average Fortune 500 enterprise will run more than 150,000 agents by 2028, up from fewer than 15 in 2025.

Governance written as policy can't keep up with that. A policy doc doesn't stop an agent that calls a tool 400 times a minute. It doesn't know which agent made the call, on whose behalf, or whether the server on the other end was ever approved. Governance breaks at the runtime, where the calls actually happen.

Here's our argument. Agent governance, identity governance, and MCP governance look like three disciplines with three owners and three toolsets. They're one problem: who's allowed to do what, through which tool, and can you prove it afterward? One problem needs one enforcement point.

What is unified AI governance?

Unified AI governance is a single runtime control layer that enforces identity, access, policy, and audit across every AI call — LLM requests, MCP tool calls, agent-to-agent traffic, and traditional APIs — using the same authenticated principal and the same policies. Policy stops being a document and becomes something every request has to pass.

The rest of this post covers what each discipline means, why separate stacks leave gaps, why identity is the key every policy runs on, what a single control layer looks like, how Kong AI Gateway implements it, and a rollout plan you can start this quarter.

What agent, identity, and MCP governance each mean

Three terms get used interchangeably. They aren't the same thing, and the differences explain why most programs end up with three separate stacks. Each definition below stands on its own.

Agent governance

Agent governance is the set of policies, permissions, lifecycle rules, and oversight that control what autonomous agents can do, including how they communicate with each other over agent-to-agent (A2A) protocols.

It differs from traditional AI or model governance in one way that matters. Models produce output. Agents take actions: they book, buy, delete, and deploy. Model governance asks whether the answer is safe. Agent governance asks whether this thing should be allowed to do that at all.

Gartner expects that by 2027, 40% of enterprises will demote or decommission autonomous agents because of governance gaps found after production incidents. It also warns that applying the same controls to every agent leads to failure, which is the case for [risk-tiered agent governance](https://www.gartner.com/en/newsroom/press-releases/2026-05-26-gartner-says-applying-uniform-governance-across-ai-agents-will-lead-to-enterprise-ai-agent-failure)risk-tiered agent governance. An agent that summarizes meeting notes and an agent that issues refunds shouldn't carry the same controls.

Identity governance for non-human identities

Identity governance and administration (IGA) manages the lifecycle, access reviews, and entitlements of the people in an organization. For AI, the useful version applies the same discipline to non-human identities: every agent gets a unique identity, a named owner, least-privilege access, and an audit trail.

Quick note on IAM versus IGA, since people search for it. Identity and access management (IAM) answers "who are you, and can you come in?" at the moment of access. IGA answers "should you still have this access, and who approved it?" over time. Agents stretch both. They need to authenticate on every call, and they need owners and reviews, because no HR system offboards an agent when the project ends.

Standards bodies are catching up. NIST's National Cybersecurity Center of Excellence (NCCoE) published a [NIST agent identity concept paper](https://www.nist.gov/news-events/news/2026/02/new-concept-paper-identity-and-authority-software-agents)NIST agent identity concept paper in February 2026 on identity and authorization for software and AI agents.

MCP governance

Search for "MCP governance" and you'll find the governance model of the open-source Model Context Protocol project itself. We mean something else: enterprise control over MCP servers and the tool calls agents make through them.

In practice that covers four areas: vetting servers before agents can reach them, authenticating and authorizing every caller, filtering which tools each caller can see, and keeping an audit record of every invocation. Our [MCP governance strategy](https://www.konghq.com/blog/enterprise/mcp-governance-agentic-ai-strategy)MCP governance strategy goes deeper on how this connects to non-human identity.

Why governing them separately creates gaps

Separate stacks mean three policy engines, three log formats, and no single record of which identity called which tool through which agent. Every system does its job. Nobody sees the whole path.

Picture a support-ticket triage agent. To get it working fast, a developer gives it a shared service account credential. A few weeks later, someone points it at an MCP server a teammate spun up for a side project. It's in no registry and no one reviewed it. That's shadow MCP, shadow IT for agent tools. The server exposes a tool that updates customer records, and the agent calls it.

Now run the incident review:

  • The identity provider (IdP) logged a token issued to a shared service account. It can't say which agent used it.
  • The agent framework logged a tool call. It doesn't know whether the server was approved.
  • The MCP server logged a request from a service account. It doesn't know which agent or person was behind it.

Three systems, three partial views, and no answer to the question the auditor asks: who changed that customer record, and who allowed it?

That one agent hits several entries in the [OWASP MCP Top 10](https://owasp.org/www-project-mcp-top-10/)OWASP MCP Top 10. The shared credential is MCP01, token mismanagement and secret exposure. A tool with more reach than the agent needed is MCP02, privilege escalation via scope creep. And a server that can't tie a request to a caller is MCP07, insufficient authentication and authorization. None of these is exotic. They're what happens when identity, tools, and agents are governed in different places.

The volume makes it worse. Gartner projects that by 2029, at least 70% of organizations running production agentic AI in infrastructure and operations will have a material service, security, or cost incident tied to agents. Its framing is [governance that goes beyond policies](https://www.gartner.com/en/articles/agentic-ai-infrastructure-governance)governance that goes beyond policies: rules enforced where the traffic flows.

Identity is the key every policy runs on

Every rate limit, tool permission, cost attribution, and audit record should key on one thing: an authenticated principal. A principal is whoever or whatever is making the call, whether that's a human, an agent, or a workload. Without one, the other policies have nothing to attach to. A rate limit on an anonymous caller is a rate limit on everyone.

The IdP stays the source of truth for who exists and what they can do. The control layer's job is to enforce that on every call, instead of trusting a token that was checked once at login.

Most organizations aren't there yet. The [CSA agentic governance maturity model](https://labs.cloudsecurityalliance.org/agentic/agentic-governance-maturity-model-v1)CSA agentic governance maturity model found that only 23% of organizations had a formal agent identity strategy, and 84% couldn't pass a compliance audit focused on agent behavior or access controls. Gartner's guidance points the same direction: treat every agent as a distinct non-human identity.

Machine-to-machine vs. user-delegated access

Agents reach tools in two ways. The difference changes how you scope, consent, and audit them.

Most production setups use both. In either case, an action that can't be traced to an identity someone's accountable for is a finding waiting to happen.

Enterprise-managed authorization for MCP

Per-user, per-server OAuth consent doesn't scale. Take 5,000 employees and 40 MCP servers. That's up to 200,000 consent screens and a pile of tokens nobody can review.

Enterprise-managed authorization moves the decision to the admin. Access is defined once, by group and role, in the IdP. The mechanism is the Identity Assertion JWT Authorization Grant (ID-JAG): a short-lived token in which the IdP vouches for a specific user requesting access to a specific resource. The user signs in through single sign-on, the MCP client trades that sign-in for an ID-JAG, and the MCP server's authorization server exchanges the ID-JAG for an access token. No consent screen. Okta Cross App Access is the first supported IdP.

Many MCP clients and servers don't support this natively yet. A gateway in the middle can relay the exchange for them, so admins get one policy surface today and the ecosystem catches up on its own schedule. For a hands-on walkthrough, see how to [secure agent traffic with OIDC](https://developer.konghq.com/ai-gateway/secure-ai-agent-with-oidc/)secure agent traffic with OIDC.

What a single AI governance control layer looks like

A single control layer puts one runtime in the path of LLM, MCP, and A2A traffic, with one control plane configuring it and shared policy, identity, and observability underneath. Here's the vendor-neutral version.

Traffic types sit on top: LLM requests to model providers, MCP tool calls, A2A messages between agents, and traditional API calls. Four governance layers apply to all of them: auth and access control, reliability, cost control, and observability and audit. One runtime enforces them. One control plane configures them.

Every traffic type passes through the same four layers. The control plane configures them, and the identity provider defines who's allowed.

Where do oversight tools fit? Inventories, risk registers, and approval workflows decide what should be allowed. Runtime enforcement makes sure it is. You need both. Oversight without enforcement is a policy doc. Enforcement without oversight is a lot of rules nobody's reviewed. Our [AI governance tools guide](https://www.konghq.com/blog/enterprise/ai-governance-tools-and-platforms-guide)AI governance tools guide breaks down which category does what.

Govern discovery: registries and tool filtering

Agents should see only approved MCP servers and tools. Start by registering every approved server with ownership metadata and a lifecycle state, such as proposed, approved, or deprecated. Publish per-audience views, so the finance agent's catalog isn't the support agent's catalog.

Then filter at discovery. A tool the caller isn't allowed to use should be invisible. If an agent never sees a tool, it can't call it, and a prompt injection can't talk it into trying.

A [governed MCP server registry](https://konghq.com/products/mcp-registry)governed MCP server registry also answers the question security teams keep asking: how do we vet MCP servers before agents use them? Nothing is reachable until someone has reviewed it and registered it.

Govern execution: authN/Z, guardrails, and cost

Every call gets checked, not just the first one. Per-call authentication and authorization confirm who's calling and whether this principal can use this tool right now. Prompt and PII guardrails inspect what goes in and out. Token quotas cap spend per principal, and chargeback attributes cost to the team that owns the agent.

Scale the controls to the risk. A read-only agent searching internal docs gets light controls: authentication, rate limits, and logging. An agent that moves money gets all of that plus tighter scopes, mandatory guardrails, and approval for high-impact actions. Locked down versus fully trusted is a false choice. For the mechanics, see [how an MCP gateway works](https://konghq.com/blog/learning-center/what-is-a-mcp-gateway)how an MCP gateway works, including IdP integration.

Govern evidence: audit and observability

If you can't reconstruct what happened, you didn't govern it. Every tool call and A2A call should produce a record with these fields:

  • Caller identity, including the delegating user when there is one
  • Agent
  • Tool or capability invoked
  • Arguments
  • Policy decision: allowed, denied, or filtered
  • Outcome
  • Latency
  • Tokens consumed

Export those records through OpenTelemetry to your SIEM, so security teams query agent activity in the same place as everything else. NIST's guidance adds a point worth building in from day one: record context and permissions, not just actions. "The agent called delete_record" is an event. "The support agent called delete_record on Maria's behalf, under a role that allowed it" is evidence.

Our page on [agent-to-agent governance](https://konghq.com/solutions/agent-gateway)agent-to-agent governance shows how caller identity lands in the audit trail for A2A traffic too.

How Kong AI Gateway unifies agent, identity, and MCP governance

The [Kong AI Gateway control layer](https://konghq.com/products/kong-ai-gateway)Kong AI Gateway control layer is the architecture above, built as one runtime and one Konnect control plane for LLM, MCP, and A2A traffic. Here's how each requirement maps to a capability.

Here’s the result: when an agent calls an MCP tool, the same authenticated principal drives access control, rate limits, and cost attribution, and the same call lands in your audit trail. No second policy engine or another log format.

If you already run Kong Gateway for API traffic, you don't need a second stack for agents. AI Gateway extends the runtime you have, so traditional APIs, LLM requests, MCP calls, and A2A traffic share one set of policies instead of four.

One boundary worth stating plainly. Kong isn't an IGA platform, an identity provider, or an agent framework. Your IdP keeps defining who exists and what they can access. Your IGA tooling keeps running access reviews. Your teams keep building agents in whatever framework they like. Kong sits in the traffic path and enforces what those systems define.

A phased rollout plan for unified AI governance

You don't need to govern everything on day one. You need to see everything first, then tighten in order. Seven steps:

  1. Inventory. List every agent, MCP server, and model in use, including the shadow MCP servers nobody registered. Ask teams, then check the traffic. The traffic is more honest.
  2. Establish agent identities. Create each agent as a non-human identity in your IdP, with a named owner.
  3. Route all AI traffic through one control layer. LLM requests, MCP calls, and A2A messages all go through the same runtime.
  4. Publish approved tools through a registry. Agents discover only what's been reviewed and registered.
  5. Apply tiered policies. Light controls for read-only agents, strict controls for agents that change data or move money.
  6. Wire audit into the SIEM. Export per-call records so security teams investigate agent activity with the tools they already use.
  7. Measure. Track the numbers below and review them on a schedule.

Metrics that show governance is working

Most governance programs can't say whether they're working. These four give you something to report.

Who owns what

IT, security, or compliance? All three, with clear lines. Platform teams run the control layer, the registry, and routing. Security and IAM teams set identity policy, create agent identities, and define risk tiers. Compliance consumes the audit evidence and sets retention and reporting requirements. Each team runs the part it's best placed to run.

Govern agents where they act

Agent governance, identity governance, and MCP governance are one problem with one answer: identity-driven governance, enforced at a single runtime. When every call carries an authenticated principal and passes through the same control layer, attribution, least privilege, and audit evidence come standard.

Speed and governance are both existential. A single control layer lets you have both. The teams that can answer "which agent, which identity, which tool" in minutes are the teams that can say yes to the next agent.

[Request a demo to explore Kong's API platform capabilities](https://konghq.com/contact-sales)Request a demo to explore Kong's API platform capabilities.

FAQs

What is MCP governance and why do enterprises need it?

MCP governance is enterprise control over MCP servers and the tool calls agents make through them. It's separate from the governance of the open-source Model Context Protocol project itself. Enterprises need it because an agent can call any tool it can reach. The main risks are shadow MCP servers nobody reviewed, over-broad tool access, and missing audit trails. Governance covers four things: vetting servers, authenticating and authorizing callers, filtering tools, and auditing every call.

How is agentic AI governance different from traditional AI governance?

Agentic AI governance covers systems that take actions, not just systems that generate output. Traditional AI governance reviews models, training data, and outputs, usually before deployment. Agents add runtime risk. They call tools, hand work to other agents, and act at machine speed, so controls have to run on every call. Lifecycle and identity requirements differ too: each agent needs its own identity, an owner, scoped permissions, and a retirement plan.

How do enterprise security teams manage AI agent identity and permissions?

Security teams give every agent a unique non-human identity with a named owner and least-privilege permissions. The IdP stays the source of truth, and a control layer enforces that identity and those permissions on every call, so access can't outlive its approval. Audit records tie each action to the agent and, when there is one, the user it acted for.

Which platforms govern A2A, MCP, LLM, and API traffic in a single runtime?

Look for platforms that run all four traffic types through one runtime and one control plane, instead of a separate gateway per protocol. Four criteria separate real consolidation from a bundle of products: shared authentication, policies keyed on the same principal across traffic types, per-call audit records in one format, and an MCP registry with tool filtering.

Kong AI Gateway takes this approach, with one runtime and one Konnect control plane for LLM, MCP, and A2A traffic, and IdP-defined access enforced on every call.

What is the difference between IAM and IGA, and where do AI agents fit?

Identity and access management (IAM) controls access at the moment of use: who you are and whether you can get in. Identity governance and administration (IGA) governs access over time: who approved it, whether it's still needed, and whether it's compliant. AI agents extend both. They need authentication and authorization on every call, which is IAM. They also need owners, entitlement reviews, and retirement, which is IGA, because nothing else offboards them.

What audit data should be captured for each MCP tool call?

Capture the caller identity (including the delegating user, if any), the agent, the tool or capability invoked, the arguments, the policy decision, the outcome, the latency, and the tokens consumed. Export the records through OpenTelemetry so they reach your SIEM in a standard format and security teams can query agent activity alongside everything else.

What's the difference between machine-to-machine and user-delegated access?

In machine-to-machine access, the agent acts as itself with its own credentials. It fits background jobs and workflows with no human in the loop, and its owner or an admin sets its scope. In user-delegated access, the agent acts on behalf of a person with that person's permissions or fewer. It fits assistants working in someone's inbox or CRM, and it needs consent from the user or an admin. Delegated access should never exceed what the user could do, and the audit record should name both the agent and the user.

Who should own AI agent governance: IT, security, or compliance?

All three, with clear lines. Platform teams run the control layer, the registry, and routing. Security and IAM teams set identity policy and define risk tiers. Compliance consumes the audit evidence and sets retention and reporting requirements.

How are MCP servers vetted before deployment?

Vet MCP servers by registering them before any agent can reach them. Review who owns each server, which tools it exposes, what authentication it supports, and what data it touches. Then publish only approved servers through a registry with a lifecycle state. Unregistered servers stay unreachable, and tool filtering shows each approved tool only to the callers allowed to use it.

How do you monitor an AI agent in production?

Monitor agents at the traffic layer, where every call passes through. Capture a per-call record of identity, tool, policy decision, outcome, latency, and tokens, and export it through OpenTelemetry to your SIEM. Then track a few numbers: unapproved tool calls blocked, token spend per principal, and mean time to attribute an incident.

How does AI change identity governance?

AI adds identities that no HR system creates or retires. Agents act at machine speed, sometimes on behalf of users, and multiply fast. Identity governance has to cover unique non-human identities, named owners, least-privilege access, and enforcement on every call. Lifecycle rules and access reviews apply to agents, not only to people.

- [Agentic AI](/blog/tag/agentic-ai)Agentic AI- [AI Gateway](/blog/tag/ai-gateway)AI Gateway

## More on this topic

_Videos_

## How Kong’s AI Gateway Governs Autonomous Agents for EU Payment Compliance

_Videos_

## Unified Control Plane for Enterprise AI Agent Governance

## See Kong in action

Accelerate deployments, reduce vulnerabilities, and gain real-time visibility. 

[Get a Demo](/contact-sales)Get a Demo
**Topics**
- [Agentic AI](/blog/tag/agentic-ai)Agentic AI- [AI Gateway](/blog/tag/ai-gateway)AI Gateway
Kong

Recommended posts

# How to Manage and Secure AI Coding Agents

[Enterprise](/blog/tag)EnterpriseOctober 8, 2026

An AI coding agent is software that writes, tests, and ships code independently. This level of autonomy earlier tools never had. It doesn't wait for you to accept a suggestion line by line. It takes a goal, breaks it into steps, and executes. To do

Kong

# AI Token Cost Management: Why AI Spend Gets Out of Control (and How To Fix It)

[Enterprise](/blog/tag)EnterpriseSeptember 24, 2026

This post is based on Kong's webinar on token cost management. Watch it below, or keep reading for the breakdown. Token cost management is how a business tracks, controls, and governs what it spends on AI model usage, the same way it already track

Kong

# AI Governance Tools and Platforms: An Enterprise Guide

[Enterprise](/blog/tag)EnterpriseSeptember 23, 2026

AI adoption is outpacing the speed at which many organizations can establish standardized oversight. In fact, Stanford HAI reported that 78% of surveyed organizations were using AI in 2024, a significant jump from 55% the previous year ( AI Index Re

Kong

# How AI Agents Communicate: Managing Context in Multi-Agent Workflows

[Enterprise](/blog/tag)EnterpriseAugust 6, 2026

An AI agent isn't magic. It's a reasoning engine that makes decisions based on what it knows at a given moment. That "what it knows" is its context — the information available to it right now. Context is everything. Give an agent the wrong context,

Hugo Guerrero

# A New Dawn: Enterprise AI's Shadow — Trillions of Tokens, Zero Governance

[Enterprise](/blog/tag)EnterpriseAugust 6, 2026

You Can't Govern What You Can't See AI spending will reach $2.59 trillion in 2026. I regularly like to share what we're seeing in production at Kong. Not projections or analyst forecasts, but actual traffic flowing through Kong AI Gateway from

Augusto Marietti

# Stop Patching. Start Building: The Kong Context Mesh Stack

[Enterprise](/blog/tag)EnterpriseJuly 23, 2026

Your infrastructure already has the raw materials: compute (VMs, containers, serverless), event streaming (Kafka, Kinesis, Pub/Sub, RabbitMQ), data stores (warehouses, databases, object storage), and AI endpoints (any hosted or self-hosted LLM). Tho

Hugo Guerrero

# From iPaaS to Context Mesh: The Architecture Shift Agentic AI Demands

[Enterprise](/blog/tag)EnterpriseJuly 23, 2026

iPaaS — Integration Platform as a Service — was built for a world of deterministic applications. The mental model is simple: you have systems, they have APIs, and iPaaS connects them. Data flows from point A to point B according to rules you define.

Hugo Guerrero

# How to Manage and Secure AI Coding Agents

[Enterprise](/blog/tag)EnterpriseOctober 8, 2026

An AI coding agent is software that writes, tests, and ships code independently. This level of autonomy earlier tools never had. It doesn't wait for you to accept a suggestion line by line. It takes a goal, breaks it into steps, and executes. To do

Kong

# AI Token Cost Management: Why AI Spend Gets Out of Control (and How To Fix It)

[Enterprise](/blog/tag)EnterpriseSeptember 24, 2026

This post is based on Kong's webinar on token cost management. Watch it below, or keep reading for the breakdown. Token cost management is how a business tracks, controls, and governs what it spends on AI model usage, the same way it already track

Kong

# AI Governance Tools and Platforms: An Enterprise Guide

[Enterprise](/blog/tag)EnterpriseSeptember 23, 2026

AI adoption is outpacing the speed at which many organizations can establish standardized oversight. In fact, Stanford HAI reported that 78% of surveyed organizations were using AI in 2024, a significant jump from 55% the previous year ( AI Index Re

Kong

# How AI Agents Communicate: Managing Context in Multi-Agent Workflows

[Enterprise](/blog/tag)EnterpriseAugust 6, 2026

An AI agent isn't magic. It's a reasoning engine that makes decisions based on what it knows at a given moment. That "what it knows" is its context — the information available to it right now. Context is everything. Give an agent the wrong context,

Hugo Guerrero

# A New Dawn: Enterprise AI's Shadow — Trillions of Tokens, Zero Governance

[Enterprise](/blog/tag)EnterpriseAugust 6, 2026

You Can't Govern What You Can't See AI spending will reach $2.59 trillion in 2026. I regularly like to share what we're seeing in production at Kong. Not projections or analyst forecasts, but actual traffic flowing through Kong AI Gateway from

Augusto Marietti

# Stop Patching. Start Building: The Kong Context Mesh Stack

[Enterprise](/blog/tag)EnterpriseJuly 23, 2026

Your infrastructure already has the raw materials: compute (VMs, containers, serverless), event streaming (Kafka, Kinesis, Pub/Sub, RabbitMQ), data stores (warehouses, databases, object storage), and AI endpoints (any hosted or self-hosted LLM). Tho

Hugo Guerrero

# From iPaaS to Context Mesh: The Architecture Shift Agentic AI Demands

[Enterprise](/blog/tag)EnterpriseJuly 23, 2026

iPaaS — Integration Platform as a Service — was built for a world of deterministic applications. The mental model is simple: you have systems, they have APIs, and iPaaS connects them. Data flows from point A to point B according to rules you define.

Hugo Guerrero

# How to Manage and Secure AI Coding Agents

[Enterprise](/blog/tag)EnterpriseOctober 8, 2026

An AI coding agent is software that writes, tests, and ships code independently. This level of autonomy earlier tools never had. It doesn't wait for you to accept a suggestion line by line. It takes a goal, breaks it into steps, and executes. To do

Kong

# AI Token Cost Management: Why AI Spend Gets Out of Control (and How To Fix It)

[Enterprise](/blog/tag)EnterpriseSeptember 24, 2026

This post is based on Kong's webinar on token cost management. Watch it below, or keep reading for the breakdown. Token cost management is how a business tracks, controls, and governs what it spends on AI model usage, the same way it already track

Kong

# AI Governance Tools and Platforms: An Enterprise Guide

[Enterprise](/blog/tag)EnterpriseSeptember 23, 2026

AI adoption is outpacing the speed at which many organizations can establish standardized oversight. In fact, Stanford HAI reported that 78% of surveyed organizations were using AI in 2024, a significant jump from 55% the previous year ( AI Index Re

Kong

# How AI Agents Communicate: Managing Context in Multi-Agent Workflows

[Enterprise](/blog/tag)EnterpriseAugust 6, 2026

An AI agent isn't magic. It's a reasoning engine that makes decisions based on what it knows at a given moment. That "what it knows" is its context — the information available to it right now. Context is everything. Give an agent the wrong context,

Hugo Guerrero

# A New Dawn: Enterprise AI's Shadow — Trillions of Tokens, Zero Governance

[Enterprise](/blog/tag)EnterpriseAugust 6, 2026

You Can't Govern What You Can't See AI spending will reach $2.59 trillion in 2026. I regularly like to share what we're seeing in production at Kong. Not projections or analyst forecasts, but actual traffic flowing through Kong AI Gateway from

Augusto Marietti

# Stop Patching. Start Building: The Kong Context Mesh Stack

[Enterprise](/blog/tag)EnterpriseJuly 23, 2026

Your infrastructure already has the raw materials: compute (VMs, containers, serverless), event streaming (Kafka, Kinesis, Pub/Sub, RabbitMQ), data stores (warehouses, databases, object storage), and AI endpoints (any hosted or self-hosted LLM). Tho

Hugo Guerrero

# From iPaaS to Context Mesh: The Architecture Shift Agentic AI Demands

[Enterprise](/blog/tag)EnterpriseJuly 23, 2026

iPaaS — Integration Platform as a Service — was built for a world of deterministic applications. The mental model is simple: you have systems, they have APIs, and iPaaS connects them. Data flows from point A to point B according to rules you define.

Hugo Guerrero

## Ready to see Kong in action?

Get a personalized walkthrough of Kong's platform tailored to your architecture, use cases, and scale requirements.

[Get a Demo](/contact-sales)Get a Demo
## step-0
    • Company
    • [About Kong ](/company)About Kong
    • [Customers ](/customer-stories)Customers
    • [Careers ](/company/careers)Careers
    • [Press ](/company/press-room)Press
    • [Events ](/events)Events
    • [Contact ](/company/contact-us)Contact
    • [Pricing ](/pricing)Pricing
      •    * [Terms](/legal/terms-of-use)
      •    * [Privacy](/legal/privacy-policy)
      •    * [Trust and Compliance](https://trust.konghq.com/)
    • Platform
    • [Kong AI Gateway ](/products/kong-ai-gateway)Kong AI Gateway
    • [Kong Konnect ](/products/kong-konnect)Kong Konnect
    • [Kong Gateway ](/products/kong-gateway)Kong Gateway
    • [Kong Event Gateway ](/products/event-gateway)Kong Event Gateway
    • [Kong Insomnia ](/products/kong-insomnia)Kong Insomnia
    • [Documentation ](https://developer.konghq.com)Documentation
    • [Book Demo ](/contact-sales)Book Demo
    • Compare
    • [AI Gateway Alternatives ](/performance-comparison/ai-gateway-alternatives)AI Gateway Alternatives
    • [Kong vs Apigee ](/performance-comparison/kong-vs-apigee)Kong vs Apigee
    • [Kong vs AWS ](/performance-comparison/kong-vsaws)Kong vs AWS
    • [Kong vs IBM ](/performance-comparison/ibm-api-connect-vs-kong)Kong vs IBM
    • [Kong vs Mulesoft ](/performance-comparison/kong-vs-mulesoft)Kong vs Mulesoft
    • [Kong vs Postman ](/performance-comparison/kong-vs-postman)Kong vs Postman
    • [Kong vs LiteLLM ](/performance-comparison/kong-vs-litellm)Kong vs LiteLLM
    • Explore More
    • [Kong for Startups ](/solutions/startup-program)Kong for Startups
    • [Open Banking API Solutions ](/solutions/open-banking)Open Banking API Solutions
    • [API Governance Solutions ](/solutions/api-governance)API Governance Solutions
    • [Istio API Gateway Integration ](/solutions/istio-gateway)Istio API Gateway Integration
    • [Kubernetes API Management ](/solutions/build-on-kubernetes)Kubernetes API Management
    • [API Gateway: Build vs Buy ](/campaign/secure-api-scalability)API Gateway: Build vs Buy
    • Open Source
    • [Kong Gateway ](https://developer.konghq.com/gateway/install/)Kong Gateway
    • [Kuma ](https://kuma.io/)Kuma
    • [Insomnia ](https://insomnia.rest/)Insomnia
    • [Kong Community ](/community)Kong Community

Kong enables the connectivity layer for the agentic era – securely connecting, governing, and monetizing APIs and AI tokens across any model or cloud.

  • English
  • Japanese
  • Frenchcoming soon
  • Spanishcoming soon
  • Germancoming soon
[Everything is 200 OK](https://status.konghq.com/)
© Kong Inc. 2026
Interaction mode