Identity broker for AI agents with OAuth2 delegation and consent management
-
Updated
Nov 16, 2026 - Go
Identity broker for AI agents with OAuth2 delegation and consent management
Kubernetes Gateway API implementation in Rust
Agent Gateway by Simplaix
MCP_HACK//26 — Industrial Zero-Trust AI: T-IA Connect + AgentGateway for Siemens TIA Portal
The AI that helps you take care of your production. Always there, never in your place.
The AI access layer for enterprises: a self-hosted LLM and MCP gateway on Kubernetes, built on agentgateway and ToolHive.
Free, hands-on course for building, securing, evaluating, and operating AI agents in Go with Google ADK and an OSS AgentOps stack.
Company-defined Markdown policies for AI gateways and coding agents. Powered by Jev, with LiteLLM, Agentgateway and GitHub Copilot connectors.
Reproducible benchmark for TypeSafe AI's Jev on agent tool-call risk classification: accuracy, latency, and whether the confidence score is worth routing on.
Tested OPA/Rego policy packs that decide whether an AI agent may make an MCP tool call: filesystem, GitHub and Postgres, default deny, every refusal explained.
The Two-Layer Agent Contract: guidance influences candidate behaviour, runtime policy enforces executable authority. Reference implementation for 'Intent Is Not Policy' talk.
Hands-on labs for securing AI agents and agentic protocols based open standards
An AI agent in a Kubernetes cluster with no route to the internet: six self-hosted models evaluated daily, everything carried in as one signed file, signed weights with an AI-BOM checked at admission, tools only through MCP, one inference gateway. A reference architecture, tested on kind.
An agent that calls an LLM without ever holding an LLM credential. Short-lived identity tokens in, provider credentials attached by agentgateway on the way out.
Seven MCP server misbehaviours run against no gateway and five enforcement setups built from OPA with a definition pin, token exchange with RLS and a typed adapter. Outcomes are read from the target; reproducible with ./run.sh.
Kubernetes-native platform for running agents you can prove things about: content-addressed revision identity, immutable release material, and rollback to the evaluated bytes.
Serve markdown to AI agents at the gateway, without touching the origin. agentgateway Accept-header content negotiation, with the two failure modes that make it silently wrong.
Hands-on demo of agentgateway v1.2.1 on kind — LLM routing, Kubernetes MCP server, content guardrails, JWT RBAC, API key auth, and prompt enrichment. Includes all manifests and setup guide.
To associate your repository with the agentgateway topic, visit your repo's landing page and select "manage topics."