ProtocolProduction ReadyA2AMCPOPENAI-FUNCTIONS

yomo

Name: yomo
Rating: 3.2 (1882 reviews)
Author: yomorun

by yomorun

QUIC-based A2A/MCP runtime for geo-distributed, low-latency agent communication

Updated Feb 3, 2026

1.9k

Stars

141

Forks

View on GitHub

Overview

Implements a serverless, geo-distributed runtime for AI agents that communicates over low-latency protocols. Uses QUIC-based transport and an A2A/MCP-inspired messaging layer to route real-time agent interactions and function-calls across edge nodes. Includes SDK and runtime primitives in Go for stream processing, function-calling, and distributed orchestration at the edge. A2A Protocol and Model Context Protocol (MCP) enable low-latency, cross-node communications.

See related protocols

Standards this tool supports

Why It Matters

As agents move to distributed, latency-sensitive deployments, a lightweight, protocol-level runtime matters for trust and evaluation: you need reproducible, observable agent interactions across regions. Yomo exposes the communication primitives and routing patterns that let you capture agent interaction logs, latency profiles, and function-call traces—signals that feed reputation and A2A evaluation. Until now most agent frameworks focused on orchestration at a single cloud region; Yomo makes geo-distributed agent communications and their failure modes first-class. Explore patterns like the Event-Driven Agent Pattern to understand reactive flows. Event-Driven Agent Pattern

Ideal For

Teams building latency-sensitive, geo-distributed agent systems that need a production-grade A2A transport and function-calling runtime. For architectural patterns that support robust routing and service mesh, consider the Agent Service Mesh Pattern. Agent Service Mesh Pattern

How It's Used

Run low-latency agent-to-agent interactions across regions for real-time coordination
Capture function-call and stream traces at the transport layer for agent evaluation
Deploy serverless agent components to edge nodes to reduce latency and observe failure modes