agents-md-sync centralizes global AI agent instructions in one canonical AGENTS.md file and installs symlinks or copies to the paths expected by common agent CLIs.

Supported agents

Agent	Target path
Amp	~/.config/AGENTS.md
Codex CLI	~/.codex/AGENTS.md
OpenCode	~/.config/opencode/AGENTS.md
Gemini CLI	~/.gemini/GEMINI.md
GitHub Copilot CLI	~/.copilot/copilot-instructions.md
Claude Code	~/.claude/CLAUDE.md

Install

uvx agents-md-sync --help

For local development:

uv sync
uv run agents-md-sync --help

Quickstart

Create the canonical source from a specific file:

uvx agents-md-sync init --from ./my-instructions.md --yes

Or let agents-md-sync init scan existing agent instruction files and ask which one to use as the base:

uvx agents-md-sync init

Install the configured source to every enabled target:

uvx agents-md-sync install --yes

Use copies instead of symlinks:

uvx agents-md-sync install --mode copy --yes

Show current target state:

uvx agents-md-sync status

Git-backed source

You can keep your canonical AGENTS.md in a Git repository. During init, pass a repo URL directly:

uvx agents-md-sync init --repo git@github.com:user/agent-config.git --dest ~/.agent-config --yes

Or select one of your GitHub repositories interactively:

uvx agents-md-sync init --select-repo --yes

The interactive selector uses gh repo list and requires the GitHub CLI (gh) to be installed and authenticated. If gh is unavailable or cannot list repos, agents-md-sync falls back to asking for the repo URL manually.

When a repo is configured, agents-md-sync init clones it into --dest (default: ~/.agent-config) or runs git pull --ff-only if the destination already exists. If the repo does not contain AGENTS.md, agents-md-sync writes the initial source file there. The config is updated with:

[source]
path = "~/.agent-config/AGENTS.md"

[repo]
url = "git@github.com:user/agent-config.git"
dest = "~/.agent-config"

After editing the configured source:

uvx agents-md-sync edit
uvx agents-md-sync push -m "update team agent instructions"

agents-md-sync push commits and pushes changes to the configured repo. It only tracks AGENTS.md, runs git pull --rebase --autostash first, and prints No changes to push when there is nothing to publish. Use --dry-run to see the commands without changing the repo.

agents-md-sync sync remains pull-only. Use it to clone or update a repository containing AGENTS.md, save it in config, and reinstall targets:

uvx agents-md-sync sync --repo git@github.com:user/agent-config.git --dest ~/.agent-config --yes

After the first sync, this is enough:

uvx agents-md-sync sync --yes

Recommended Git-backed workflow:

uvx agents-md-sync init --select-repo
uvx agents-md-sync edit
uvx agents-md-sync push
uvx agents-md-sync sync --yes

Configuration

By default, config is stored at ~/.config/agents-md-sync/config.toml:

[source]
path = "~/.config/AGENTS.md"

[[targets]]
name = "my-agent"
path = "~/.my-agent/instructions.md"
enabled = true

Ad-hoc targets are also supported:

uvx agents-md-sync install --extra-target my-agent=~/.my-agent/instructions.md --only my-agent --yes

Development

This project uses uv, ruff, ty, pytest, and pytest-cov.

uv sync
uv run ruff check .
uv run ruff format --check .
uv run ty check
uv run pytest

uv run pytest enforces 100% coverage for agents_md_sync.

No pre-commit setup is required.