## AI Cost Management: Attribute, budget, measure, & optimize
Agent spend grows quietly. Usage shows up as an item line on a shared cloud or model invoice, long after the agents that caused it have run, and most teams cannot tell which team, harness, or app drove it.
AI Cost Management solves this with brand new cost governance solutions:
- - **Dynamic price catalog**: a price catalog keeps provider token and spend rates in one place, with the ability to define custom overrides for custom provider deals.
- - **Flexible attribution engine**: design attribution models to track spend by team, harness and app
- - **Native integrations with different third-party AI tools**: track and attribute spend against different coding harnesses, copilots, SDLC tools, and more
- - **Budget definition**: Set budgets against attribution groups and define hard and soft budgets to stop overruns before they reach the invoice
- - **A new cost management dashboard**: built on all of the above and the Gateway visibility into token consumption, you can view and break Down spend against budget ceilings, different teams, harnesses, repos
If you want to learn more, check out [the deep-dive blog on AI Cost Management](https://konghq.com/blog/product-releases/kong-ai-cost-management)the deep-dive blog on AI Cost Management.
## Token Vault: Keep multi-system credentials out of agent code and prompts
Agents need enterprise credentials to do useful work, and today they often hold them directly, in code, configuration or prompts. Every agent that holds a credential is another place it can leak. The problem grows with MCP: each server provider supports its own auth flow, some use OAuth with user consent, others use API keys, some leverage enterprise IdP on-behalf-of (OBO) flows, service accounts, etc. So when an agent accesses multiple MCP servers—internal and external—access control becomes a nightmare for developers and a major liability for security teams. When, not if, an agent behaves unexpectedly, controlling the blast radius of credential exposure is critical. Kong Token Vault enforces the principle of least privilege, preserves end-to-end auditability, and enables immediate, targeted credential revocation.
Token Vault solves these issues by ensuring:
- - **Credentials stay out of agents**: the agent never holds a token, so no agent code, configuration, or prompt contains one.
- - **Agent MCP access can be brokered across providers: **multi-system credentials can be vaulted in Token Vault, and then Token Vault + the Gateway handle credential exchange and enforcing access control against those
[Learn more >](https://konghq.com/blog/product-releases/kong-token-vault)Learn more >
## Context Mesh: Turn existing APIs and MCP servers into governed agent tools, once
Enterprises already have APIs and MCP servers, but making them truly agent-ready — which means they must be token- and context-efficient — usually means either wrapping each one by hand (in the case of APIs) or rewriting existing MCP servers to support methods such as code mode.
Context Mesh solves this and generates governed MCP servers from the APIs and MCP servers you already run:
- - **Governed MCP servers from existing APIs**: each service becomes an agent tool, with policy applied once instead of in every copy.
- - **Composites**: combine several sources across APIs or MCP servers into one governed MCP server.
- - **Catalog and AI Registry**: one place to find and reuse APIs, MCP servers and AI assets across teams.
The generated servers use code mode. Instead of loading a definition for every tool into the agent's context, the agent writes short code that calls only the tools it needs. That keeps token use and context size down as the estate grows, so agents can work across many APIs and MCP servers without filling their context with tool definitions.
And of course, all of these MCP servers can be immediately deployed to a Kong AI Gateway for full governance.
[Learn more >](https://konghq.com/blog/product-releases/context-mesh)Learn more >
## AI Observability: See what every agent did
When an agent misbehaves or an audit asks what it did, request metrics are not enough. AI traffic has to be read as sessions, steps and model calls, and most tooling was built for API requests.
AI Observability reads AI traffic as sessions, steps and model calls, through three views:
- - **Dashboard**: an overview of AI traffic.
- - **Explorer**: query AI traffic across the dimensions that matter to your teams.
- - **Logs**: the individual AI requests behind every number.
- - **Advanced AI Observability** (coming in late November or early December): session-and-turn-level tracing across agent steps, tool calls and model calls, so teams can explain and audit what an agent did.
[Learn more >](https://konghq.com/blog/product-releases/kong-advanced-ai-observability)Learn more >
## The best news: It all comes with the Gateway
Context Mesh, Token Vault, AI Cost Management, and Basic Observability are included with Kong AI Gateway on Konnect, with no separate SKU. Teams already running the gateway in Konnect can use these capabilities without buying anything extra. If you want to learn more, reach out to your Kong point of contact or [book a demo](https://konghq.com/contact-sales)book a demo.
**Topics**
Alex Drag
Head of Product Marketing
# Stop Patching. Start Building: The Kong Context Mesh Stack
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
# The Control Tower Reaches into the Runtime: Announcing the ServiceNow and Kong Partnership
Enterprise AI has moved from pilots to production faster than most governance programs could keep up. Agents now run across every major cloud and model provider, built by central teams, business units, and developers alike. Each one reasons with mod
Alex Drag
# The Architecture Decision Your Multi-Agent System Will Live With
Multi-agent systems are, at their core, context distribution systems. Every agent in your workflow is a consumer and producer of context. The interesting architectural questions are all about how that context moves. Two operations drive everything:
Hugo Guerrero
# Kong AI Gateway 2.0 Is Now GA
If you're running agents in production, you've felt this problem: every agent, every LLM application, every MCP client needs to be configured against a growing sprawl of MCP servers, each with its own endpoint, its own handshake, and its own access
Greg Peranich
# Why Your AI Agents Keep Failing (Hint: It's Not the Model)
Over the last few years, enterprises have made a genuinely impressive leap. They moved from rule-based automation to large language models capable of reasoning, planning, and acting. The intelligence layer is dramatically better. But the integration
Hugo Guerrero
# A Unified Gateway for APIs + Agentic Applications on VMware VKS with Kong Konnect
Built on top of Kong API Gateway, the Kong AI Gateway is designed to address key challenges in enterprise AI adoption. Modern AI applications rarely rely on a single model; instead, they orchestrate multiple GenAI providers, agent frameworks, Age
Anika Suri
# AI Agent Integration: Gartner Research Confirms Need for AI Control Layer
An AI control layer is the governance and observability infrastructure that sits between AI agents and enterprise applications, handling authentication, routing, rate limiting, and auditability to ensure secure, managed access. Unlike traditional in
Heather Halenbeck
