On November 4, 2026, Kong introduced AI Management in Kong Konnect, adding controls for AI spending, credentials, governed discovery and observability for production agents. Cost management, credential handling, context and basic observability come with Kong AI Gateway on Konnect rather than as separate add-ons. The move matters because multi-model agent tasks now cross models, MCP servers, APIs and tools, making scattered controls hard to sustain.
Cost, credentials and discovery in one gateway
Kong frames the release around four operational gaps: spending that cannot be assigned to a team or application, credentials embedded in agent code or prompts, APIs repeatedly wrapped as agent tools, and no consolidated account of agent activity. The company places the controls in the connectivity path so policy and usage data follow interactions across models, MCP servers and agents. Governed discovery for models, MCP servers and agents is part of the broader release. Existing gateway customers on Konnect receive the core components without another purchase.
AI Cost Management links gateway consumption data with request identity and context to attribute spending to business units, teams, applications, agents, users, customers or tenants. A price catalog centralizes provider token and spending rates and permits overrides for negotiated arrangements. Integrations with third-party AI tools extend attribution to coding harnesses, copilots and development workflows. Teams can set budgets for those groups and receive alerts as consumption approaches thresholds. The Konnect overview adds hard and soft budget options plus a dashboard comparing spending with ceilings by team, harness or repository.
Optimization operates at the infrastructure layer through intelligent model routing, semantic caching, prompt caching, prompt compression and reasoning controls. Kong notes the benefit varies by workload, with semantic caching suited to repetitive requests and routing useful when expensive models handle tasks cheaper alternatives could perform. A forthcoming Opportunities feature for finding efficiencies and measuring savings is marked as coming soon. Token Vault, launched with per-user three-legged OAuth in Kong AI Gateway 2.3, is included with all Kong Gateway subscriptions. Context Management had already become generally available on September 30, 2026.
What this means for production AI operations
For companies running copilots and coding harnesses, attribution down to team, repository, user or tenant changes how AI bills are managed and charged back. Hard budgets can cap consumption while soft budgets warn before limits are hit, and the dashboard shows where expensive model use concentrates. Small firms gain a packaged way to track spending without building metering, while large organizations can align AI costs with existing cost-center logic. Routing and caching offer a path to lower token bills without rewriting applications.
Token Vault removes upstream credentials from agent code by authenticating users and agents to an MCP server proxy with corporate identity-provider credentials. On first use the gateway returns a short-lived enrollment link, the user completes normal OAuth consent, and refresh tokens are encrypted with the customer key for storage in Kong Identity. Later calls decrypt the credential in memory using a private key from the customer secrets manager, insert it upstream and refresh it as needed. The agent sees the response without seeing the token, and removing policy access takes effect at once even before upstream revocation.
The marker to watch is delivery of advanced session-and-turn tracing, scheduled for late November or early December 2026. That workflow should expose inputs, outputs, tokens, cost, latency, policy checks and tool calls for an individual turn. Its arrival, alongside the Opportunities savings measurement, will show whether Konnect functions as a shared accountability layer for agents. Until then, core cost, identity and discovery controls are available, while full audit detail remains pending.
