AgentMesh Documentation

Composable multi-agent orchestration for Go

AgentMesh helps teams build production AI systems with streaming outputs, deterministic control, and pluggable observability.

Why AgentMesh?

  • ✅ Deterministic flows (Sequential, Parallel, Loop)
  • ⚙️ Strongly typed tool integration with JSON Schema
  • 📡 Streaming-first event pipeline with partial updates
  • 🧩 Pluggable stores for session, memory, artifact data
  • 🔭 Built-in logging, tracing, and metrics hooks

What is AgentMesh?

AgentMesh is a Go framework for composing reliable AI agents. It gives you deterministic orchestration patterns, streaming event pipelines, and pluggable observability so you can ship production-grade AI systems with confidence.

Key ideas:

  • Compose agents as sequential pipelines, parallel fan-outs, or iterative loops.
  • Expose strongly typed tools (including MCP providers) to your models.
  • Stream partial and final results to downstream consumers.
  • Plug in your logging, metrics, tracing, and storage without forking the runtime.

Core capabilities

  • Agent patterns: compose sequential, parallel, and looping agents to model complex workflows.
  • Tool ecosystem: register strongly typed tools with JSON Schema validation, plus ship retrievers and toolsets with shared observability.
  • Streaming orchestration: flows emit partial and final events in real time for responsive UX.
  • Observability: integrate structured logging, metrics, and tracing via pluggable providers.
  • Stateful sessions: manage history, artifacts, and memory through configurable stores.
  • Extensibility: intercept model and agent lifecycles with plugins for custom behavior.

Explore the documentation

  • Getting started → – Install the module, run the quick start, and browse local tooling tips.
  • Models guide → – Connect OpenAI, LangChainGo, or custom providers with structured outputs and tool calls.
  • Agents guide → – Learn how Sequential, Parallel, Loop, Model, and Func agents compose orchestration graphs.
  • Tools guide → – Build function tools, the AgentTool wrapper, and MCP-backed toolsets with confidence.
  • Plugins guide → – Hook into runner, agent, model, and tool lifecycles with reusable interceptors.
  • Observability → – Configure logging, metrics, and tracing providers and read them from context.
  • Architecture → – Understand the flow engine, runner lifecycle, and how the core packages fit together.
  • Resources → – Jump to examples, issues, and contribution guides.