A live replica with sample content. Click the island to try it yourself.
How it works
- 1
Connect Claude Code or Codex from Eflow. Eflow adds one hook entry to that tool's own settings; nothing else about the tool changes.
- 2
While an agent works, the closed island shows what it is doing, step by step.
- 3
When it needs permission, to run a command for example, the island opens and shows exactly what it wants to do and in which project.
- 4
Approve or reject with one click, and the agent carries on or stops right away.
- 5
Type a follow-up to that same session from the island, or switch between your Claude Code and Codex sessions.
- 6
When the work is done, a pill on the island says so.
Good to know
Works with Claude Code and Codex.
Your prompts and the agents' replies go directly between you and Anthropic or OpenAI, exactly as they would without Eflow. Eflow shows what is happening and passes on your answer.
If Eflow isn't running, the agent simply asks you in its own window, as usual.
Codex asks you to trust a new hook once, with its /hooks command.
You can hide agent progress from the closed island in Settings.
Questions
Which AI coding agents does Eflow work with?
Claude Code and Codex.
How does Eflow connect to Claude Code or Codex?
It adds one hook entry to the tool's own settings, so its permission requests and progress can be shown on the island. Nothing else about the tool changes.
What happens if Eflow is closed?
The agent asks for permission in its own window, exactly as it would without Eflow.
Does Eflow send my code or prompts anywhere?
No. Eflow shows the agent's steps and requests on your PC and sends none of it anywhere. Your prompts and the agent's replies travel between you and Anthropic or OpenAI as usual.