shenli/awesome-agent-infra

Awesome Agent Infra is a curated, organized list of links about the infrastructure needed to run AI agents safely and reliably in production. An AI agent, in this context, is software that uses a language model in a loop to take actions, like calling tools, editing files, or completing multi-step tasks on its own. This project is not about the agents themselves or about prompt writing. Instead it focuses on the layer of systems that sit underneath agent products and above the raw AI models, things like the software that runs an agent's loop, the isolated environments where its actions actually execute, and the tools that let it discover and call outside services. The list is organized into clear categories: runtime and control planes, tool protocols like MCP, workspace and runtime state, sandboxes and execution environments, security and policy, observability and evaluation, and coding-agent workflows, among others. It also includes a glossary of core concepts, explaining terms like agent runtime, harness, workspace, sandbox, and publish lifecycle in plain definitions, along with links to foundational articles from companies such as Anthropic, OpenAI, Temporal, and GitHub that explain these ideas in more depth. The project is explicit about what it leaves out. It does not include general prompt collections, generic chatbot UI tools, broad AI app directories, or typical retrieval-augmented generation tools unless they connect directly to agent infrastructure. This keeps the list focused on the systems layer rather than becoming a catch-all for anything AI-related. This resource is best suited for developers, engineers, and technical leaders who are building or evaluating tools to run AI agents in real production environments, rather than casual users experimenting with a single chatbot. It functions purely as a reference and reading list, not as code you install or run yourself. The full README is longer than what was shown.