Someone on your team ran a Claude Code session that got most of the way to solving a problem. You want to continue from there without re-reading the entire thread or starting cold. That is what forking is for. This guide covers how to fork a Claude Code or Codex session with Lore, what a fork carries forward, and when forking beats starting fresh.
The short answer
Open the shared session in Lore, copy its fork command, and run it in your own coding agent:
/fork {threadId} I want to...
You fill in the intent after the thread ID. Lore generates an AI-distilled handoff prompt from the original session, conditioned on what you said you want to do next, and your agent starts from that context instead of from nothing.
What forking actually does
A fork is not a copy of the transcript. It is a handoff. When you fork a thread, Lore reads the source session, extracts the key decisions and context, and distills them into a compact prompt (usually a short list of bullet points) tuned to your stated intent. That distilled prompt is what you feed your own Claude Code or Codex session.
The result: your agent begins with an understanding of what the original session established, what it decided, and where it left off, without you pasting in thousands of lines or re-deriving the reasoning yourself.
This is why the intent matters. "I want to add tests for the parser this session built" produces a different handoff than "I want to port this approach to the Go service." The distillation is conditioned on what you are trying to do.
How to fork, step by step
- Open the session in Lore. It can be your own thread or one a teammate shared with you (workspace or public visibility).
- Find the fork command. The session view shows a command in the form
/fork {threadId} I want to... with a copy button.
- Add your intent. Replace
I want to... with a short description of what you want to do next. Be specific about the goal.
- Run it in your agent. Paste the command into your own Claude Code or Codex session. Your agent picks up the distilled context and continues from there.
When to fork vs. when to start fresh
Fork when there is real context worth carrying forward:
- A teammate solved a hard problem and you need to extend or adapt it.
- A long session established constraints and decisions you would otherwise have to rediscover.
- You are moving an approach from one part of the codebase to another and want the reasoning to come with it.
Start fresh when the task is genuinely new, or when the source session was short enough that reading it directly is faster than a handoff. Forking earns its keep on sessions where the reasoning is dense and re-deriving it would cost you real time.
Fork vs. share
Sharing and forking solve adjacent problems. Sharing (/share) makes a session readable: you send a URL, and someone opens it to see what happened. Forking (/fork) makes a session continuable: you take its context into a new session to do more work.
You often do both. Share the session so the team can see it, then fork it when someone is ready to build on it.
Frequently asked questions
What does it mean to fork a Claude Code session?
Forking means continuing from where a session left off. With Lore, forking generates an AI-distilled handoff prompt from the original session, conditioned on what you want to do next, which you run in your own Claude Code or Codex agent so it starts with that context.
Does forking copy the whole transcript into my session?
No. A fork is a distillation, not a copy. Lore extracts the key decisions and context from the source thread and compresses them into a short handoff prompt, so your agent starts informed without ingesting the entire transcript.
Can I fork a session a teammate shared with me?
Yes, as long as you can open the thread (it is shared with your workspace or made public). Copy the fork command from the session view, add your intent, and run it in your own agent.
Can I fork a Codex session too?
Yes. Once a Codex session is shared to Lore with /share-codex, it behaves like any other thread, including forking. The fork command works the same way regardless of whether the source was Claude Code or Codex.
How is forking different from just re-running the prompts?
Re-running prompts replays instructions without the reasoning or the decisions made along the way. A fork carries forward a distilled understanding of what the session established and where it ended, tuned to your intent, so you continue the work rather than repeat it.
A practical starter
The next time a teammate shares a session that gets close to something you need, do not start over. Open it in Lore, write a one-line intent for what you want to do next, and fork it. You will spend the first minute building on their work instead of reconstructing it.