I built Portable Handoff because I was getting annoyed at how much context disappeared between coding sessions. After spending an hour with Claude Code or Codex, the agent would finally understand the project, what I had already tried, the constraints, and what needed to happen next. Then the session would end, I would switch tools, or the context would fill up, and I had to explain everything again. I tried using normal summaries first. They were useful, but I found it hard to tell what was actually true versus what the model remembered or inferred. That led to Portable Handoff. The model records things like decisions, blockers and next steps, while the local tool records facts such as the Git commit, branch, file hashes and timestamps. I mainly wanted something I could use myself when moving between Claude Code, Codex and Cursor without restarting the project from scratch. It is open source and still early, so I would be very interested to hear how other people handle this problem today.

Portable-handoff

Stop explaining your project every time you switch AI tools.

Stop explaining your project every time you switch AI tools.