UNIVERSAL GATEWAY · PROTOCOLS
Nine protocols. One policy plane.
Agent systems no longer speak a single protocol. The Universal Gateway terminates the ones that matter for AI workloads at one edge — so authentication, rate limits, logging, and policy are written once and enforced everywhere.
What the gateway speaks.
MCP — Model Context Protocol
Used by AI assistants and agent frameworks to call tools and fetch context. The gateway brokers MCP so each tool server sits behind auth, quotas, and audit instead of being wired directly into the model.
A2A — Agent-to-Agent
Used when autonomous agents delegate work to one another. The gateway gives every hop an identity and a policy check, so a chain of agents stays attributable end to end.
HTTP / REST
Used by virtually every service, SDK, and webhook. It is the universal default: point any REST upstream at the gateway and inherit routing, transforms, and observability for free.
MQTT
Used by IoT fleets, sensors, and edge devices that publish telemetry. The gateway bridges MQTT pub/sub into the same control plane as your request traffic for unified policy and logging.
WebSocket
Used by live agent sessions, streaming token output, and real-time dashboards. The gateway proxies full-duplex connections with auth and per-connection limits intact.
gRPC
Used by performance-sensitive service-to-service backends. The gateway handles gRPC routing and health checking for low-latency internal calls without losing observability.
GraphQL
Used by data-rich agent front-ends that want typed, client-shaped queries. The gateway fronts GraphQL endpoints with the same auth and rate-limiting as the rest of the surface.
AG-UI
Used to render interactive agent user-interfaces driven by the backend. The gateway carries AG-UI traffic so agent surfaces are governed like any other protocol.
ACP — Agent Communication Protocol
Used for interoperable messaging between agents across vendors and runtimes. The gateway terminates ACP so cross-system agent messaging is consistent and auditable.
Why terminate them all in one place.
Every protocol you add to an agent system is another place to get authentication, rate limiting, and logging wrong. Spread across application code, the controls drift apart and an audit becomes archaeology.
Terminating all of them at the Universal Gateway means a single policy definition covers MCP tool calls, A2A delegations, and HTTP webhooks alike. One edge, one audit trail, one place to answer the question a regulator actually asks: what is each agent allowed to do, and can you prove it?
FAQ
Frequently asked questions.
- What is MCP (Model Context Protocol)?
- MCP is the protocol AI assistants and agent frameworks use to call tools and fetch context from external systems. The Universal Gateway brokers MCP traffic behind authentication, quotas, and audit so tool access is governed rather than open.
- What is the A2A (Agent-to-Agent) protocol?
- A2A is used when autonomous agents delegate work to one another. The Universal Gateway gives every agent-to-agent hop an identity and a policy check, so delegation chains stay observable and auditable instead of opaque.
- Why route every protocol through one gateway?
- Each protocol is another place to get authentication, rate limiting, and logging wrong. Terminating MCP, A2A, HTTP, MQTT, WebSocket, gRPC, GraphQL, AG-UI, and ACP at one edge means policy is defined once and enforced everywhere, producing a single audit trail across the entire agent surface.
Bring your protocols under one control point.
Tell us which protocols your agents speak today and where the policy lives. We will map them to a single gateway configuration in a scoped pilot.