agent-sync

Keep Claude Code and Codex configs in sync without rewriting everything by hand.

agent-sync keeps the shared parts shared, generates the tool-specific files you still need, and tells you what needs manual follow-up.

Quick Start

For now, run it from this repo:

git clone git@github.com:matanabudy/agent-sync.git
cd agent-sync
uv sync

Start with a dry run to see the analysis:

agent-sync sync --dry-run .

Then apply it:

agent-sync sync .

Fixing Warnings With Claude or Codex

Some parts of the migration are intentionally reported instead of rewritten. Use the generated report to clean those up with Claude or Codex.

After sync, open:

• .agent-sync/sync-report.md
• .agent-sync/sync-report.json

Then ask Claude or Codex to fix the remaining portability issues in your repo.

For example:

Read .agent-sync/sync-report.md and fix the portability issues one by one.
Keep behavior the same where possible and explain any warnings that need manual follow-up.

What Migrates

Surface	Quality	Notes
Instructions	High	Plain markdown maps cleanly
Skills	High	Name, description, body, support files
MCP servers	Medium	Stdio and HTTP map cleanly
Agents	Medium-low	Structural mapping, not 1:1
Permissions / sandboxing	Low	Advisory only

Not migrated cleanly, but detected and reported:

• Claude hooks, plugins, auto-memory, @path imports, .claude/rules/
• Codex .rules execution policy, profiles, tool-specific execution settings

For more details, see docs/claude-codex-migration-analysis.md.

Development

From this repo:

uv sync
uv run pytest
uv run ruff check src tests
uv run ty check