Follow a question without losing the main thread.

What is branching AI chat?

In a single long chat, questions about examples, definitions, and alternatives all sit in one sequence. A branch gives a question its own path. You can explore it, return to the original explanation, and follow another question from there.

Grenvi anchors a branch to selected text. The tree and map show how the conversations connect, so you can return to a particular line of thought.

Try your first branch

1. Open Grenvi and ask about a topic you are studying or researching.
2. Select a sentence in the response that you want to understand better. Choose the action to create a new branch.
3. Ask your follow-up in the new conversation. You can keep asking questions along that path.
4. Use the return navigation or the conversation tree to go back. The original conversation remains available.
5. Open the map to see the main conversation and the branches you explored.

What context does a branch use?

A branch follows the conversation path from the root to its current position and carries the selected passage as an anchor. Content from parallel branches is kept separate from that branch’s conversation context.

The interface helps you navigate and separate lines of inquiry. It does not guarantee a correct answer or an infinite model context window. Long conversations still depend on the product’s context handling and available AI allowance.

Choose a branch or an annotation

Use a branch when you want to investigate a new question. Use annotations when you want to revise the response in the current conversation. You can select several passages, add a request to each, and submit the annotations together.

For example, open a branch to ask why a formula works. Add an annotation to ask for a clearer definition or a different example in the original answer.

Turn exploration into a project

When the conversation has a clear direction, organize it into a project. Review the proposed focus and outline before developing connected chapters. The original exploration stays available.

Start with one real task rather than trying every feature at once. A concept you are learning or a question you are researching is enough.