Manage chats and context

Use chats to carry out tasks, discuss choices, and propose intent changes. Including relevant documents, explanations, and attachments helps the Agent understand exactly what you mean.

Start or resume a chat

New Chat opens an empty input view. The chat is saved and added to the sidebar only after you send its first message. Closing it before sending leaves no empty chat behind.

Before the first message, choose an Agent, model, reasoning effort, and permissions. Once the Agent connects, that chat cannot switch to another Agent; start a new chat instead. See Agents, models, and usage for these settings.

Opening an old chat first displays local history. The Agent connects or resumes when you send a new message.

Unsent text and attachments survive switching chats, but not restarting the app. Copy important drafts or save them in a document.

Reference documents and explanations

Type @ at the beginning of a line or after whitespace to search documents, paragraphs, or explanations. A selected item becomes a reference chip and is supplied to the Agent with your message. This identifies the target more clearly than “the paragraph above.”

You can also choose Ask about this from a paragraph's actions. It inserts a reference and focuses the input, waiting for you to write and send the question.

Enable reference previews in Settings → General to preview content on hover in the reference menu. References in sent messages remain clickable so you can return to the document or explanation.

Attach files

Click the attachment button, drag files into the input, or paste images or files. Images appear as thumbnails; other attachments appear as file chips.

Click an image to enlarge it. Use arrow keys or the navigation buttons to browse images in the same group. Remove any unwanted attachments before sending.

Whether an image can be sent directly to the model depends on the Agent. If that capability is unavailable, Noea supplies it as a file and explains this in the chat.

Use commands

Type / at the start of the input to open the current Agent's command menu. It groups Noea Skills and Agent commands. Choose an item, add any arguments, and send it.

The available commands depend on the Agent and configured Skills; they are not necessarily the same across Agents. See Add MCP servers and Skills.

Read the execution trace

The Agent's reasoning and tool calls appear in an expandable execution trace. View the current step while it runs; after completion, the trace collapses into a summary. Expand file edits or command calls for details.

Intent references in replies open documents, code files open in the system's default application, and external pages open in a browser. The sources summary at the end of each turn collects relevant references.

Edit messages and branch chats

While a chat is idle, you can edit a sent user message or branch a new chat from an Agent reply.

Submitting an edited message continues from the revised message. The original message and subsequent history remain in a branch you can switch back to. Branching from a reply copies the history through that reply into a new chat.

Neither action rolls back files the Agent changed. The new branch works with the implementation currently on disk. Restore code separately using version control or file changes if needed.

Available edit and branch points depend on Agent support. Running chats cannot be edited, branched, or switched between branches. Automatically generated task messages from Explain, organization, or proposal acceptance cannot be edited either.

Organize chat history

A chat's sidebar menu lets you rename, pin, or unpin it. Pinned chats appear under Pinned. Once you name a chat yourself, automatic title generation will not overwrite that name.

Deleting a chat requires confirmation and also deletes its local attachment files. Running chats cannot be deleted. Use search to find older discussions before deciding what to remove.

Record conclusions that should become lasting requirements by reviewing intent changes.