trpc-group/trpc-agent-go

A Go framework for building production agent systems with graph workflows, tools, memory, A2A, AG-UI, MCP, evaluation, and observability.

What it solves

Building production-grade AI agents requires managing complex workflows, persistent memory, tool integration, and observability. Standard LLM scripts often lack the concurrency, type safety, and service-oriented architecture needed for enterprise deployment.

How it works

tRPC-Agent-Go provides a Go-native runtime that treats agents as scalable services. It uses a graph-based workflow engine for complex routing, a modular memory system for session persistence, and a rich tool ecosystem (including MCP support) for external interactions. It also features a "self-evolution" mechanism where completed sessions are reviewed and converted into reusable skills.

Who it’s for

  • Go Developers building scalable, concurrent AI services.
  • AI Engineers needing robust orchestration, evaluation, and observability for multi-agent systems.
  • Enterprise Teams requiring production-ready features like OpenTelemetry tracing and persistent state management.

Highlights

  • Multi-Agent Orchestration: Supports chaining, parallel execution, and cyclic workflows.
  • GraphAgent: Type-safe graph workflows with multi-conditional routing.
  • Agent Self-Evolution: Asynchronously reviews sessions to extract and publish reusable skills.
  • Production Observability: Built-in OpenTelemetry tracing and Langfuse integration.
  • Rich Tooling: Supports function tools, MCP, web search, and code execution.
  • Memory & RAG: Integrated memory services and knowledge retrieval for context-aware agents.