> For the complete documentation index, see [llms.txt](https://docs.nimbalyst.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nimbalyst.com/session-management/agent-window-and-session-management.md).
# Agent Window & Session Management
### Agent Window
In the File Window, you focus on editing your markdown files.
In the Agent Window, you focus on your sessions with your Agent. These sessions could include long chat discussions, research sessions, complex multi-file tasks, and AI-assisted development.
The Agent Window consists of three main areas:
* Left Sidebar showing your Sessions. You can Search sessions here and select them
* Central transcript area to converse with agent with full conversation history, tool call visualization, streaming content display
* Right panel showing you files edited and read by the agent and prompts used in the conversation
### Session Management
Run multiple parallel sessions.
Manage Sessions by:
* Searching
* Filtering
* Resuming
Relate Sessions to Documents and Documents to Sessions.
### Session Features
* **Turn summary stats** — each agent turn shows file count and line changes (e.g., "Finished in 6m 57s — 3 files +45 -12")
* **Click-to-copy code blocks** — click or tap inline code blocks in transcripts to copy with visual feedback
* **Maximize button** — in the chat sidebar, click the maximize button to open the current session in full agent mode
* **Drag-drop to create Workstreams** — drag a session onto a standalone session to create a Workstream
* **Start session from a document** — "Start a new agent session" from a document pre-populates with an @file reference to that document
* **Rename a session** — right-click an existing session in the list and choose Rename. Sub-session renames live-update; workstream parent rows also support inline rename
* **Preferred Agent Language** — under **Settings > Application > Agent Features**, set the language used to auto-generate session names (applies to Claude Code, Codex, and other providers)
### Start a Background Session from Anywhere
Press **Cmd+Shift+N** from any Nimbalyst mode to open the session launcher over your current work.
Start a session without leaving the document, Tracker, pull request, or shared file you are viewing.
The launcher includes the same controls as the normal chat composer:
* Choose the agent and model.
* Set effort and supported thinking controls.
* Insert an Action or slash command.
* Mention files and Tracker items.
* Attach images and other supported context.
Drag the launcher by its title bar when it covers something you need to read. Sending the prompt starts the session in the background, closes and resets the launcher, and keeps you in the current mode. Open Agent mode or the attention list whenever you want to follow the new session.
Starting a session from a document or Tracker item can prefill the relevant reference so the agent receives the context you launched from.
### Change Models and Thinking from Chat
Press **Cmd/Ctrl+Shift+M** while the chat input is focused to open the model picker without reaching for the mouse. Type to search by the model's display name or model ID, then use the arrow keys and Enter to select it.
The model selection belongs to the session. Changing it affects later turns without changing the model used by other sessions.
Supported Claude Agent models also show **Extended: On** or **Extended: Off**:
* **Extended: On** keeps the model's extended thinking behavior.
* **Extended: Off** can reduce latency and token use when the task does not need as much reasoning.
The control appears only for models and providers that support it. Extended thinking stays on by default.
### Review the Context Sent with a Prompt
When you select content in an editor, the chat composer shows that selection as a removable chip before you send the prompt.
Selection chips can represent:
* Selected text in documents and code files
* Spreadsheet cells
* A selected mockup screen
* One or more Excalidraw shapes
* Selection types contributed by other editors
Remove a chip when you do not want that selection sent to the model. Removing it changes only the pending prompt; it does not change or deselect the content in the editor.
### Contextual Tips
Empty sessions and other core surfaces can show contextual tips instead of a blank state. These cards point you toward features that fit what you are doing right now.
Common examples include:
* Opening **Tracker Mode** when your work is spreading across many sessions
* Discovering worktree-specific flows and search shortcuts
* Learning keyboard shortcuts and launchers without leaving the current screen
* Finding shared docs, theme controls, feedback entry points, and mobile pairing
Tips often include a direct action like **Open Tracker**, plus **Next** and **All tips** controls so you can browse more guidance without interrupting your current work.
### Session Kanban Board
Open the Session Kanban with **Cmd+Shift+K** to see all sessions in the workspace grouped by phase (Backlog, Planning, Implementing, Validating, Complete).
#### Live peek transcripts
Hover any session card and the transcript opens in a peek popover. If the session is running, the transcript streams in live as the agent works, so you can keep an eye on what it is doing without opening the session itself. The peek shows the recent context of long turns, not just the trailing tokens.
* Hover to open, move away to dismiss
* Streaming sessions update token by token in the peek
* Works for sessions, workstreams, and child sessions inside a workstream
### Self-Pacing Session Wakeups
Agents can schedule themselves to wake up later and continue working — useful for long-running tasks where the agent needs to check back on something after a delay rather than running continuously.
* The agent picks the wakeup time itself (anywhere from 60 seconds to 7 days out)
* Wakeups persist across app restarts, but only fire while Nimbalyst is running
* When a wakeup fires, you get an OS notification — clicking it focuses the workspace and opens the session
* A wakeup banner appears in the session header with **Cancel** and **Fire now** controls; sessions with a pending wakeup show a clock icon in the session list
If the workspace window is closed when the wakeup is due, the wakeup waits and re-fires the next time you open that workspace.
### Agent Attention List
The Agent navigation icon shows a badge when sessions are running, unread, or waiting for you.
The attention list groups sessions by what they need, so a blocked agent does not disappear into session history.
Open the icon to see sessions grouped as:
* **Awaiting input** — the agent is blocked on a question, approval, or permission.
* **Running** — the agent is actively working.
* **Unread** — the session has new activity you have not opened.
Select a session to open it. **Mark all read** clears unread state, but it does not approve requests or answer sessions that are awaiting input.
Agents can also send a local system notification when a workflow needs your attention and you have stepped away. Clicking the notification returns you to the relevant workspace and session. Notification availability follows your operating-system and Nimbalyst notification settings.
### System Tray Icon
The system tray menu shows session status at a glance:
* See which sessions are running, completed, or need attention
* Click any session to navigate directly to it
* The dock and Agent navigation badges show a count for sessions that need your attention
### Window State Persistence
The Agentic Coding Window remembers:
* Window size and position
* Active tab in sidebar
* Session ID
* Developer tools state (if open)
* Transcript scroll position
State is saved per workspace.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://docs.nimbalyst.com/session-management/agent-window-and-session-management.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.