# Astromesh

Astromesh is an open-source runtime that standardizes how AI agents execute and interact with external systems: agents are defined declaratively in YAML, routed to different models per role, orchestrated with built-in patterns, and deployed as a REST and WebSocket API.

Rather than a team writing its own model orchestration, tool execution, memory, retrieval and observability code for every new agent project, Astromesh centralizes those pieces into one runtime and lets a developer define an agent's roles, tools and behavior declaratively instead of imperatively. Seven built-in reasoning patterns are available, including ReAct, plan-and-execute, fan-out, pipeline, supervisor and swarm styles, and a router can pick a model per role by cost, latency, quality or capability match.

Models can come from Ollama, any OpenAI-compatible API, vLLM, llama.cpp, Hugging Face TGI or ONNX Runtime, and deployments include retrieval-augmented generation, memory, MCP tool access, guardrails and OpenTelemetry observability out of the box, with several deployment modes documented alongside a full architecture reference.

It is written in Python and released under the Apache-2.0 license.

Website: https://monaccode.github.io/astromesh/