agentgateway v1.3.0-alpha.1 drops today.
A2A backends, Okta MCP auth, resource multiplexing, ext authz caching, dtrace upgrades, and a ton of fixes.
Big step forward for agentic proxying. https://t.co/w7dRA6rqOf
agentgateway v1.2.1 is out — small but important patch release.
Fixes CRD installs on older Kubernetes + improves MCP handling and load balancing.
Full details: https://t.co/ouonWlBsKH
Your agents shouldn’t manage ten MCP connections. They should manage one.
https://t.co/gJEodXRfHY federates GitHub, Jira, Slack, Sentry and more behind a single /mcp endpoint — every tool, one connection, unified routing.
#MCP#AIAgents#Kubernetes#agentgateway#DevTools
3 security implementations you need from your AI Gateway:
🟢 Standard Auth (OIDC-based oAuth) and OBO
🟢 Guardrails, rate limiting, prompt guards, and any other name for "should this Agent be allowed to do what it's doing?" (a higher-level name for this is Policy Enforcement)
🟢 Observability for EVERYTHING that goes through your gateway. This could be for anything from token count and usage to traces for ensuring you know exactly what an Agent is doing.
There are many other things as well (flexibility, scalability, traffic management, etc.), but these are the top three implementations that a gateway should have for managing agentic traffic in production.
#agenticai #ai
I hope this is true
BREAKING: Google has said it will REFUSE to comply with Carney's Surveillance Bill C-22, by refusing to provide encryption keys to CSIS and the RCMP.