Build AI agents you can understand and maintain

Pick a starting point

The site is organized around problems, not product names. Start with the topic closest to the system you are trying to build or troubleshoot.

Agent orchestrationUnderstand what an agent runtime coordinates, where skills and tools fit, and when multiple agents help.

Memory architectureSeparate live context from durable notes, searchable indexes, and lossless conversation history.

MCP toolingLearn how Model Context Protocol servers expose tools and context, and when an API or CLI is the better interface.

Infrastructure and securityApply the same ideas to homelab services, secrets, deployment decisions, and safe automation.

Terms you will see here

• OpenClaw is an agent runtime and gateway. It connects models to workspaces, tools, skills, sessions, and communication channels.
• LCM, short for Lossless Context Management, is the approach used by the Lossless Claw plugin to compact long conversations without discarding the underlying messages.
• QMD, short for Query Markup Documents, is a local search engine for Markdown and other text collections. It combines keyword search, vector search, and reranking, and it can expose search through MCP.
• MCP, the Model Context Protocol, is a standard connection between an AI application and servers that expose tools, resources, or prompts.

Each guide defines specialized terms again where they matter. You should not need prior experience with a specific agent framework to follow the architecture. Use the agent systems glossary when you need a definition without the full implementation guide.