Go Micro

Go Micro is an open-source Go framework for building microservices designed to be orchestrated by AI agents. Originally created in 2015, the project has evolved to v5 and now centers on AI-native microservice development, where every service endpoint is automatically exposed as an AI-callable tool via the Model Context Protocol (MCP). The repository reports over 22,000 GitHub stars and is licensed under Apache 2.0.

What It Is

Go Micro is a production-ready microservices framework for Go that bridges traditional distributed systems patterns with modern AI agent workflows. Developers write services as ordinary Go structs with methods; the framework handles registration, discovery, RPC communication, and pub/sub events. The key differentiator in v5 is that every endpoint automatically becomes an MCP tool, allowing LLMs to discover and call services without any additional wiring. An micro chat command lets an LLM orchestrate across all running services as a unified agent.

AI-Native Architecture

The framework's AI layer sits on top of its core distributed systems primitives:

• MCP gateway: Every service endpoint is exposed as an AI tool at http://localhost:8080/mcp/tools automatically.
• Agent orchestration: micro chat connects to a configured LLM provider and lets it discover services from the registry and call them as tools.
• Service generation: micro run --prompt "..." sends a plain-English description to an LLM, which designs the architecture, generates Go handlers with real business logic, compiles them, and starts the services.
• Mid-conversation generation: When the agent needs a capability that doesn't exist, it can generate a new service on the fly during a chat session.
• Seven LLM providers: Anthropic (claude-sonnet-4), OpenAI (gpt-4o), Google Gemini, Groq, Mistral, Together AI, and Atlas Cloud are supported with a unified interface.

Core Framework Capabilities

Beneath the AI layer, Go Micro provides a full microservices toolkit:

• Service discovery: mDNS by default, with Consul and etcd as alternatives.
• RPC and streaming: gRPC transport with load balancing and streaming support.
• Pub/sub messaging: NATS, RabbitMQ, and HTTP broker built in.
• Typed data model: CRUD and query layer with SQLite, Postgres, and in-memory backends.
• Hot reload: micro run watches files and rebuilds on change.
• Templates: micro new --template crud/pubsub/api scaffolds common patterns.
• Deployment: micro deploy user@server pushes via SSH and systemd with no Docker required.
• Pluggable interfaces: All abstractions are Go interfaces, so any component can be swapped without changing service code.

Developer Workflow

The typical workflow starts with either micro new to scaffold a service from a template or micro run --prompt to generate one from a description. Running micro run starts a local environment with a dashboard at http://localhost:8080, a REST/gRPC API gateway, an agent playground, and MCP tool endpoints. micro chat then connects an LLM to the running services. For production, micro deploy handles SSH-based deployment with systemd process management. Generated code can be edited by hand at any time; re-running preserves manual changes.

Update: v5.25.0

The latest release is v5.25.0, published on June 3, 2026, with the repository last pushed on June 4, 2026. The v5 line represents a significant evolution from earlier versions, adding the full AI/MCP layer on top of the established microservices core. The project lists Anthropic and Atlas Cloud as sponsors on the homepage and in the README.