- CollectionLLM Gateway & Routing
- Classificationagent-native
- MCPAvailable
- Agent SkillUnavailable
- InterfacesMCP · A2A · STDIO · http · API
What it does
Agentgateway is an open-source proxy for agentic traffic: it routes agent-to-LLM requests through an OpenAI-compatible surface (multi-provider), connects agent-to-tool via MCP (including federation and auth), and agent-to-agent via the A2A protocol. It adds JWT/API key auth, CEL-based RBAC, rate limits, prompt guards, and OpenTelemetry visibility — positioned as infrastructure for agent workloads rather than a generic HTTP reverse proxy.
Agent-facing interfaces
- Interaction pattern: self-hosted proxy (binary, Docker, or Kubernetes) — OpenAI-compatible LLM front door plus MCP and A2A routing
- Catalog protocols: MCP, A2A, STDIO, http, API
- MCP: ✅ Core protocol (MCP gateway is a primary feature)
- Agent Skill: ⚠️ No official SKILL.md / npx skills add … pack for Agentgateway yet.
- Onboarding type:
mcp
How agents onboard
Interaction pattern: self-hosted proxy (binary, Docker, or Kubernetes) — OpenAI-compatible LLM front door plus MCP and A2A routing
curl https://raw.githubusercontent.com/agentgateway/agentgateway/refs/heads/main/common/scripts/get-agentgateway | bash
curl -sL https://raw.githubusercontent.com/agentgateway/agentgateway/main/examples/basic/config.yaml -o config.yaml
agentgateway -f config.yaml
# Open UI at localhost:15000
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| OpenAI-compatible HTTP | Chat completions proxy to many providers |
| MCP | stdio, SSE, Streamable HTTP — client and server federation |
| A2A | Agent-to-agent protocol routing |
| Kubernetes | Gateway API / control-plane deployment options |
| Configuration | YAML locally; CRDs on Kubernetes |