tRPC-Agent-Go
A Go framework for building production agent systems, providing LLM agents, graph-based workflows, tool calling, session and memory state, knowledge retrieval, evaluation and OpenTelemetry observability in one Go-native stack.
tRPC-Agent-Go is a Go framework for building production agent systems. It provides a Go-native agent runtime with streaming, context cancellation and service-friendly APIs, a type-safe GraphAgent for multi-conditional graph workflows functionally equivalent to LangGraph, and multi-agent collaboration through chain, parallel and cycle-based workflows.
It ships a tool ecosystem covering function tools, MCP tools, web search and code execution, persistent session, memory and artifact state with knowledge retrieval, reusable SKILL.md agent skills with safe execution, and session reviews that extract, gate and publish new skills. It integrates AG-UI for front ends, A2A for agent interoperability, and MCP for tools, with prompt caching for cost savings, evaluation sets and metrics for tracking quality over time, and OpenTelemetry tracing and metrics with example integrations.
It targets use cases such as customer support bots, data analysis assistants, DevOps automation and RAG-powered research agents, and is written in Go under the Apache 2.0 license.