Kong Releases AI Gateway 2.2 With MCP Tool Access Control
Kong K.K. began general availability of Kong AI Gateway 2.2, an update that adds governance and cost controls for enterprise AI agents.
1 source just now
Reporting from 1 source: ASCII.jp.
Why it matters
Kong is positioning the gateway as the control point for agent traffic, and the same release ships the identity, credential and cost plumbing that make per-agent policy enforcement possible.
Key Facts
- Kong Inc. announced Kong AI Gateway 2.2 on October 1, 2026, and Kong K.K. began its general availability on Kong Konnect.
- Kong AI Gateway 2.2 adds MCP Server Bundling, which consolidates multiple MCP servers into a single Kong route with per-user and per-agent tool permissions.
- Kong AI Gateway 2.2 adds native AWS IAM SigV4 authentication for AI agents and MCP servers running on Amazon Bedrock AgentCore.
- Kong AI Gateway 2.2 adds native support for Kimi, Microsoft Foundry, Amazon SageMaker and the decision-focused model Jev.
- Kong's AI Cost Management feature tracks AI spend by media type and cache reads and writes and supports Headroom prompt compression.
The release, announced by US parent Kong Inc. on October 1, 2026, is available on the Kong Konnect platform without a beta enrollment step. Its headline feature, MCP Server Bundling, merges multiple MCP servers into one Kong route so only the tools permitted for a given user or agent are exposed and executed. The update also adds cost tracking split by media type (text, audio, image, video) and by cache reads and writes, plus support for Headroom prompt compression. Kong Identity supplies rate limiting, usage analytics, cost allocation and access control keyed to authenticated user and agent IDs. Native AWS IAM SigV4 authentication covers agents and MCP servers on Amazon Bedrock AgentCore, and the gateway adds native support for Kimi, Microsoft Foundry, Amazon SageMaker and the decision-focused model Jev. Senior Vice President of Product Reza Shafii said agentic AI adoption depends on enterprises being able to use agents with confidence.
Kong CEO and co-founder Augusto Marietti framed the update around agent behavior rather than model access. "For AI agents to deliver value, they must be able to securely connect with APIs, tools, data, applications, and events," he said. "And all of these interactions need to be discovered, authorized, governed, monitored, and cost-managed."
Shafii's statement went further than the summary version, calling the release "one of our most important evolutions to date" and tying governance to three needs: visibility into how AI is operating, control over what agents can access and execute, and management of risk and cost.
The cost side of the release rests on a broader set of features than per-media-type tracking alone:
- Cost calculation by model and provider pricing, accounting for input and output tokens, cached tokens and service tiers.
- Cost allocation across users, agents, teams, applications, products and cost centers.
- Budget planning aligned to organizational priorities.
- Usage forecasting and anomaly detection, with policies applied when usage runs abnormal.
- Optimization through semantic caching, prompt compression and intelligent model routing, with the savings measured.
Kong cites a Gartner prediction that by 2028, enterprises that do not measure AI token usage will reach 70% less of their AI-related growth targets. The company also points to showback and chargeback reporting for departmental cost allocation.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google