Make Agent Friendly: preparing codebases for agentic development

Here's a scenario that comes up constantly: you fork a repository or inherit a codebase that was built entirely by humans, and now you want to work on it with an AI coding agent. The repo works fine, but it wasn't designed with agent navigability in mind. There's no CLAUDE.md, no context directory, and the structure might not be optimal for how agents parse and understand code.

What the plugin does

Make Agent Friendly is a Claude Code plugin that guides Claude through a systematic process to prepare any repository for agentic development. When you run /make-agent-friendly, Claude analyzes your codebase, asks clarifying questions about your intended workflow, then works through a structured checklist: restructuring folders for agent navigability, creating CLAUDE.md and /context/ infrastructure, updating documentation, and validating that nothing breaks in the process.

The agentic development model

The plugin is built around a clear division of labor: humans control planning, provide instructions, and supervise, while the agent executes code changes, suggests improvements, and navigates the codebase. For this to work well, the repository needs to be structured so the agent can quickly find what it needs, understand the project context, and know where to put things.

After running the command, your repo typically ends up with a CLAUDE.md for project context, a /context/ directory for external docs and API references, optionally a /planning/ directory for decision records, and an updated README reflecting the agentic workflow.

Installation

You can install it via the plugin marketplace or manually clone and add it as a local plugin. There's also a legacy slash command option for backwards compatibility. The whole thing is customizable since the command prompt and reference documentation can be forked and modified.

GitHub: Make-Agent-Friendly.