By bendrucker
Automatically loads AGENTS.md files into Claude Code sessions to provide persistent project instructions and context.
Claude Code plugin for reading AGENTS.md files.
Loads project instructions from AGENTS.md files automatically, enabling Claude Code to work with projects using this open standard.
If you work in a small number of repositories, symlinking AGENTS.md to CLAUDE.md is probably fine:
ln -s AGENTS.md CLAUDE.md
This plugin is intended for users who work across many repositories and want consistent AGENTS.md support without per-project setup. It's also useful for large monorepos where linking AGENTS.md to CLAUDE.md in every subdirectory is harder to reliably maintain.
With the plugin installed, AGENTS.md files load automatically as you navigate your codebase. Subdirectory files take precedence over parent files, giving you hierarchical context without manual symlink management.
This plug-in tries to adhere as closely as possible to the AGENTS.md spec, which has subtle differences from CLAUDE.md.
claude plugin install github:bendrucker/claude-code-agents-md
Add the bendrucker marketplace, then install the plugin:
claude plugin marketplace add bendrucker/claude
claude plugin install agents-md@bendrucker
The plugin uses two hooks to inject AGENTS.md content into Claude Code sessions:
SessionStart: At session start (or after /clear, /compact), injects the root AGENTS.md from the project directory.
PostToolUse (Read): After any file read, walks up the directory tree from the read file to find the closest AGENTS.md and injects it. This allows subdirectories to have their own context.
Each AGENTS.md is injected only once per session to avoid context duplication.
Create AGENTS.md files in your project to provide context to Claude Code:
project/
├── AGENTS.md # Root instructions (injected at session start)
├── src/
│ ├── api/
│ │ └── AGENTS.md # API-specific instructions
│ └── utils/
└── tests/
└── AGENTS.md # Testing instructions
When Claude reads a file in src/api/, it receives the src/api/AGENTS.md content. The closest AGENTS.md to the file being read wins.
See the AGENTS.md standard for guidance on file contents.
AGENTS.md vs CLAUDE.mdClaude Code natively supports CLAUDE.md files for project instructions. This plugin adds support for AGENTS.md, an open standard.
| Behavior | CLAUDE.md | AGENTS.md |
|---|---|---|
| Loads nearest file in directory tree | ✓ | ✓ |
| Loads all ancestor files to project root | ✓ | ✗ |
@ file imports | ✓ | ✗ |
Use AGENTS.md for portable project instructions across AI coding tools. Use CLAUDE.md for richer Claude Code-specific features.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub bendrucker/claude-code-agents-md --plugin agents-mdQuery local shell history from atuin: recent commands, usage stats, and ad-hoc SQL over the history db
Pull request workflows: create, update, babysit CI, and reviewer follow-up
Query local ActivityWatch usage data: per-app time, window titles, and active vs idle spans
Meta-tools for Claude Code configuration and customization
Create technical specifications through interactive planning and expert review
Ultra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.
Memory compression system for Claude Code - persist context across sessions
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
45% cost reduction measured. Cache expiry prevention, SubTask auto-delegation, zero-cost context restoration, real-time cost dashboard. The only Claude Code plugin built from CC source analysis.
Harness-native ECC plugin for engineering teams - 67 agents, 279 skills, 94 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses