AgenticGoKit

AgenticGoKit is an open source Go framework for building multi agent AI systems. It positions itself as the Go counterpart to LangChain in Python, giving Go developers a native way to build agent workflows with dynamic tool integration, support for multiple LLM providers and orchestration suited to production services. It is aimed at teams that prefer compiled Go services over Python environments. The API centres on a small set of builder methods with compile time type checking, and agents deploy as a single binary that uses goroutines for true parallel execution. Four workflow types are built in, sequential, parallel, DAG and loop, and they can be nested as subworkflows with shared state. Memory and retrieval augmented generation use built in vector database support for pgvector and Weaviate, tools are discovered dynamically through the MCP protocol, and real time streaming comes with typed errors and retry logic. The documentation shows research assistants, conversational agents with persistent memory, data processing pipelines and RAG knowledge bases as typical builds. AgenticGoKit is free and released under the Apache 2.0 license. There is no paid tier; running costs depend on the model providers and databases you connect. The framework is currently in beta, and its APIs may change before a stable version 1.0 release. The older core and vnext APIs are deprecated, and the documentation includes a migration guide to the current v1beta interface.