AgenticGoKit
A Go framework for building multi-agent AI systems, with a streaming-first API, built-in memory and retrieval-augmented generation, and sequential, parallel, DAG, and loop orchestration patterns.
AgenticGoKit is a Go framework for building intelligent multi-agent systems, with a unified, streaming-first API covering single agents up through complex multi-agent workflows. It supports multiple LLM providers interchangeably, including OpenAI, Anthropic, Ollama, Azure OpenAI, Azure AI Foundry Local, Hugging Face, OpenRouter, vLLM, BentoML, and MLFlow, and accepts multimodal input, images, audio, and video, alongside text.
Multi-agent workflows can be composed as sequential, parallel, DAG-based, or looping orchestration patterns, and the framework ships with memory and retrieval-augmented generation built in by default with no configuration required, swappable for pgvector or a custom backend. Built-in distributed tracing over OpenTelemetry supports production observability. AgenticGoKit is the core of a three-part ecosystem that also includes a separate command-line tool for scaffolding, tracing, and evaluating agent workflows, and a template registry powering those scaffolds. The v1beta API is the current recommended interface, with an eventual v1.0 release planned to make it the primary package.