Allow editing sent agent messages and regenerating from the edited prompt

[eluzzi5]

Summary

Feature request: let users edit a message they already sent to a Warp agent, then regenerate the agent's response from the edited prompt.

Problem

When a user sends an agent prompt with typos, missing context, or unclear wording, the current workaround is to send a corrective follow-up or retype the full prompt. This adds extra effort and can clutter the conversation, especially when the original prompt was long and only a small correction is needed.

The requested workflow is moderate friction in normal agent use rather than a complete blocker.

Reproduction steps or desired workflow

1. Send a message to a Warp agent.
2. Edit the sent message from the message UI, such as a hover/context-menu action, or via a keyboard shortcut similar to editing the last command.
3. Submit the edited message.
4. Warp replaces the original message and regenerates the agent response from that point in the conversation.

A follow-up enhancement could let the user choose whether to regenerate or preserve the existing response, but the primary desired behavior is to regenerate from the edited prompt.

Artifacts

None attached.

Warp version

Unknown.

Operating system

macOS, version unknown.

[oz-for-oss]

@eluzzi5

Here's what I found while triaging this issue:

This appears to be a feature gap in Agent Mode rather than a macOS-specific bug. The current UI supports nearby actions like copying, forking, and rewinding conversations, but not editing a sent prompt and regenerating from that point.

Maintainer details

I concluded that feature request to add an Agent Mode edit-and-regenerate workflow for previously sent user prompts..

Triage

This is an Agent Mode enhancement request, not a bug report. The requested behavior is to edit a previously sent user prompt and regenerate the agent response from that edited point in the conversation.

Current code paths expose adjacent primitives: AI block menus include actions such as copying the prompt/conversation and saving the prompt, the block/header UI can open AI block overflow actions, the terminal view has fork and rewind/truncation flows, and the controller appends new user inputs to an existing conversation. I did not find an existing edit-resubmit action that replaces a prior AIAgentInput::UserQuery, truncates dependent exchanges, and starts a new response stream from the revised prompt.

Duplicate check: I found related but not exact open requests, including #6690 for broadly editable conversations and #9139 for a sticky prompt header that mentions an edit/resubmit quick action. They do not provide two strong matches for this specific edit-sent-message-and-regenerate workflow, so I am not marking this as a duplicate.

Suggested routing: Agent Mode/product UX. This likely needs design/product decisions around whether editing is allowed for only the latest user message or any prior user message, how regenerated branches/history are represented, and how tool side effects from discarded responses are handled.

This is my automated analysis and may be incorrect. A maintainer will verify the details.

Powered by Oz