agconf

CLI to sync AI agent configurations across repositories.

Documentation

• Canonical Repository Setup - Setting up a source repository
• Downstream Repository Configuration - Customizing sync behavior
• Versioning - How version management works
• File Integrity Checking - How integrity is enforced
• Contributing - Contributing guidelines

Full documentation available on GitHub: https://github.com/julian-pani/agconf

Commands

| Command | Description | Example | |---------|-------------|---------| | init | Initialize repo from a canonical source | agconf init --source org/standards | | sync | Sync content from canonical repo (fetches latest by default) | agconf sync or agconf sync --pinned | | check | Verify managed files are unchanged | agconf check | | upgrade-cli | Upgrade the CLI to latest version (auto-detects package manager) | agconf upgrade-cli | | canonical init | Scaffold a new canonical repository | agconf canonical init | | canonical update | Update CLI version in workflow files | agconf canonical update | | config show | Show current configuration | agconf config show | | completion install | Install shell completions | agconf completion install |

For detailed command documentation, see the Canonical Repository Setup and Versioning guides.

Quick Start

1. Create a canonical repository

mkdir engineering-standards && cd engineering-standards
git init
agconf canonical init --name my-standards --org "My Org"

This scaffolds the structure for your standards. Edit instructions/AGENTS.md to add your engineering guidelines, then commit and push to GitHub.

2. Sync to your projects

cd your-project
agconf init --source your-org/engineering-standards

CLAUDE.md Handling

During sync, agconf consolidates any existing CLAUDE.md files into AGENTS.md and creates .claude/CLAUDE.md with a reference to it. This ensures a single source of truth while maintaining compatibility with both Claude Code and GitHub Copilot.

Rules

Rules are modular, topic-specific project instructions synced from your canonical repository. For Claude Code, they're placed in .claude/rules/ as separate files. For GitHub Copilot, they're concatenated into AGENTS.md under a "Project Rules" section.

Rules support subdirectory nesting and can include paths frontmatter for conditional loading (Claude only).

Configuration: Add rules_dir: "rules" to your canonical agconf.yaml

For detailed information on rules setup, directory structure, and target-specific behavior, see the Rules section in Canonical Repository Setup.

Agents

Agents are Claude Code sub-agents (markdown files with YAML frontmatter) synced from your canonical repository. They define specialized AI assistants that can be invoked for specific tasks.

Target-specific behavior:

• Claude Code: Agents are copied to .claude/agents/ as flat files with metadata for change tracking
• GitHub Copilot: Not supported (Copilot does not have sub-agents)

Configuration: Add agents_dir: "agents" to your canonical agconf.yaml

Each agent file requires frontmatter with name and description fields:

---
name: code-reviewer
description: Reviews code changes for quality and best practices
---

# Code Reviewer Agent

## Instructions
...

For detailed information on agents setup and file format, see the Agents section in Canonical Repository Setup.

Downstream Configuration

Downstream repositories can optionally customize sync behavior by creating .agconf/config.yaml. This allows you to control commit strategy (direct commits vs pull requests), commit messages, and PR reviewers.

Example: Set direct commits instead of creating PRs:

workflow:
  commit_strategy: direct
  commit_message: "chore: sync engineering standards"

For complete configuration reference and available settings, see Downstream Repository Configuration.

License

MIT