tags
Click a filter to scope results to projects or posts.
MCP — the Model Context Protocol — is the standard interface for plugging external tools and data sources into AI agents. In Claude Code, MCP servers turn the CLI into a hub that can query your database, search your Notion, browse your Linear, control your browser, and run any tool you can wrap in a process.
Subagents in Claude Code are independent AI workers spawned by the main agent to handle scoped tasks — code search, planning, security review, status-line setup. They run with their own context window, their own tool allowlist, and return a single summary. This post explains when to use them, how to author your own, and the patterns that work.