Pinned post
Today we launched AI Management in Konnect, and itโs one of the most important things weโve ever shipped.
Our customers can now address all of these challenges in one place:
โช AI Cost Management
โช Agent Credentials
โช Context Management
โช Observability
The platform teams we work with kept running into these same four problems, so we built a control plane that handles all of them.
๐๐ ๐๐ผ๐๐ ๐ ๐ฎ๐ป๐ฎ๐ด๐ฒ๐บ๐ฒ๐ป๐
Agent usage shows up as a line item on a shared model invoice, long after the agents that caused it have run. AI Cost Management gives you a price catalog with custom overrides for your provider deals, attribution by team, harness, and app, and hard and soft budgets that stop overruns before they reach the invoice.
๐๐ด๐ฒ๐ป๐ ๐๐ฟ๐ฒ๐ฑ๐ฒ๐ป๐๐ถ๐ฎ๐น๐
Agents hold enterprise credentials in code, config, and prompts, and every copy is another place for a leak. MCP makes it worse. Each server brings its own auth flow, whether OAuth with user consent, API keys, IdP on-behalf-of, or service accounts. Token Vault keeps the agent from ever holding a token. The vault and the gateway broker the exchange, enforce least privilege, keep an audit trail, and let you revoke a credential instantly when an agent misbehaves.
๐๐ผ๐ป๐๐ฒ๐ ๐ ๐ ๐ฒ๐๐ต
Enterprises already run APIs and MCP servers, and making them agent-ready usually means wrapping each one manually. Context Mesh generates governed MCP servers from what you already have, with policy applied once. The servers use code mode, so the agent writes short code that calls only the tools it needs, and your context window doesnโt fill up with tool definitions.
๐๐ ๐ข๐ฏ๐๐ฒ๐ฟ๐๐ฎ๐ฏ๐ถ๐น๐ถ๐๐
Request metrics canโt explain what an agent did. AI Observability reads traffic as sessions, steps, and model calls, with session-level tracing coming soon.
Context Mesh, Token Vault, AI Cost Management, and Basic Observability all ship with Kong AI Gateway on Konnect.
Congrats to Saju Pillai, Reza Shafii , and the entire engineering and product teams for pulling this off.
Full details: konghq.com/blog/product-rโฆ
Our customers can now address all of these challenges in one place:
โช AI Cost Management
โช Agent Credentials
โช Context Management
โช Observability
The platform teams we work with kept running into these same four problems, so we built a control plane that handles all of them.
๐๐ ๐๐ผ๐๐ ๐ ๐ฎ๐ป๐ฎ๐ด๐ฒ๐บ๐ฒ๐ป๐
Agent usage shows up as a line item on a shared model invoice, long after the agents that caused it have run. AI Cost Management gives you a price catalog with custom overrides for your provider deals, attribution by team, harness, and app, and hard and soft budgets that stop overruns before they reach the invoice.
๐๐ด๐ฒ๐ป๐ ๐๐ฟ๐ฒ๐ฑ๐ฒ๐ป๐๐ถ๐ฎ๐น๐
Agents hold enterprise credentials in code, config, and prompts, and every copy is another place for a leak. MCP makes it worse. Each server brings its own auth flow, whether OAuth with user consent, API keys, IdP on-behalf-of, or service accounts. Token Vault keeps the agent from ever holding a token. The vault and the gateway broker the exchange, enforce least privilege, keep an audit trail, and let you revoke a credential instantly when an agent misbehaves.
๐๐ผ๐ป๐๐ฒ๐ ๐ ๐ ๐ฒ๐๐ต
Enterprises already run APIs and MCP servers, and making them agent-ready usually means wrapping each one manually. Context Mesh generates governed MCP servers from what you already have, with policy applied once. The servers use code mode, so the agent writes short code that calls only the tools it needs, and your context window doesnโt fill up with tool definitions.
๐๐ ๐ข๐ฏ๐๐ฒ๐ฟ๐๐ฎ๐ฏ๐ถ๐น๐ถ๐๐
Request metrics canโt explain what an agent did. AI Observability reads traffic as sessions, steps, and model calls, with session-level tracing coming soon.
Context Mesh, Token Vault, AI Cost Management, and Basic Observability all ship with Kong AI Gateway on Konnect.
Congrats to Saju Pillai, Reza Shafii , and the entire engineering and product teams for pulling this off.
Full details: konghq.com/blog/product-rโฆ
1:59