Cross-language project porting and synchronization tool powered by AI agents.

upsync automates the process of porting projects between programming languages and keeping ports synchronized with upstream changes. It uses AI coding agents (Claude Code, Gemini CLI, or Codex CLI) to translate code while maintaining behavioral equivalence.

cargo install --path .

At least one AI coding agent must be installed:

• Claude Code (claude)
• Gemini CLI (gemini)
• Codex CLI (codex)

Create a new port of an upstream project:

upsync port https://github.com/org/project --to rust
upsync port https://github.com/org/project --to go --output my-project

This will:

1. Clone the upstream repository
2. Detect the source language
3. Create a new project in the target language
4. Set up tracking for future synchronization

If you've already manually ported a project and want to track upstream changes:

cd your-ported-project
upsync init https://github.com/org/upstream-project

This will:

1. Clone the upstream repository to .upstream-cache/
2. Detect source and target languages
3. Create configuration files for sync tracking

upsync check

Fetches the latest upstream commits and reports how many lines have changed since the last sync.

upsync sync           # Apply changes
upsync sync --dry-run # Preview what would happen

The sync command:

1. Fetches upstream changes
2. Generates a diff since the last sync
3. Invokes the AI agent to apply equivalent changes
4. Runs verification to ensure the build passes
5. Updates the sync state

Created in your project root:

upstream: https://github.com/org/project
agent: claude          # Optional: claude, gemini, or codex (auto-detected if not set)
target_language: rust  # Optional: override detected language
verify_command: cargo check  # Optional: custom verification command

Tracks synchronization state (managed automatically):

{
  "last_synced_commit": "abc123...",
  "source_language": "typescript",
  "target_language": "rust"
}

Local clone of the upstream repository used for fetching changes.

1. Language Detection: Counts file extensions to determine the dominant language
2. Diff Generation: Uses git to track changes between synced commits
3. AI Translation: Sends structured prompts to AI agents with:
   • Source and target language context
   • The upstream diff (for sync) or full codebase (for port)
   • Verification commands to run
4. Retry Loop: If verification fails, the agent is re-invoked with error context (up to 3 attempts)

MIT