Originally published at terminalblog.com
Related articles
- GitHub Copilot CLI: The Terminal Agent That Knows Your Repos
- Best Coding Agents 2026 — Decision Guide
- Claude Code Now Runs Code Reviews in the Background and Gets Opus 5
GitHub Copilot CLI v1.0.76, released July 30, adds a Sessions sidebar for managing multiple concurrent coding sessions, a queue manager that lets you reorder and edit messages before they are sent, and Grok 4.5 as a new model option. These three features address one of the longest-standing complaints about terminal coding agents: the single-session bottleneck that forces you to either finish one task before starting another or juggle multiple terminal windows with no shared visibility.
If you have ever been mid-way through refactoring an API endpoint in one Copilot session and needed to quickly fix a bug in another file without losing your place, the Sessions sidebar is the feature that makes this possible. If you have ever queued a message and then realized you wanted to change its order or edit the prompt before sending, the queue manager removes the frustration of watching the wrong message go first.
Sessions sidebar: multiple coding sessions in one window
The biggest change in v1.0.76 is the Sessions sidebar, which lives on the left side of the terminal and shows all your active Copilot sessions at a glance. You can switch between sessions, spawn new ones, and see which session is doing what without leaving the current one.
To turn it on, run /experimental on in any Copilot CLI session. The sidebar appears as a collapsible panel. Each session card shows the session name, its current status (running, waiting, idle), and which model it is using. Click or use keyboard navigation to switch between sessions. The active session card is accented by default to make it obvious which one you are working in.
For beginners, think of this like browser tabs for your coding agent. Before this feature, each Copilot CLI session was isolated in its own terminal window. If you wanted to work on two tasks simultaneously, you needed two separate terminal windows, and switching between them meant losing visual context on what the other one was doing. The Sessions sidebar keeps everything in one place, showing you the status of each session so you can pick up where you left off.
What happens to your mode when you switch
A small but important detail: when you switch away from a session and come back, Copilot now restores the mode you were in. If you were running in autopilot mode — where the agent completes tasks autonomously — and you switched to another session, coming back resumes autopilot instead of dropping you into interactive mode. This means the task_complete tool stays available, and you do not have to re-enable autopilot every time you switch sessions.
Before this fix, resuming a session always reverted to interactive mode, which meant autopilot sessions would lose their autonomous behavior the moment you looked at another task. For anyone using autopilot for longer-running work, this was a silent productivity killer.
Queue manager: control the order and content of queued messages
The second major feature is a queue manager accessible through the /queue command. When you send a message while Copilot is busy processing another one, the message gets queued. Before v1.0.76, queued messages were fire-and-forget — you could not change their order, edit their content, or remove them without canceling the entire queue.
The queue manager gives you full control:
- Reorder queued messages by moving them up or down in the list
- Edit the prompt text of any queued message before it is sent
- Remove individual messages from the queue
- Repeat a message by re-queuing it after it has been sent
- Send immediately by selecting a queued message and skipping ahead in the queue
This matters because the order of prompts affects how the agent works. If you queued "run the test suite" before "fix the failing test," the agent would try to run tests before understanding what was wrong. With the queue manager, you can swap the order so the agent investigates first and runs tests second.
You can also remove your own newest queued message with Ctrl+C, which is useful when you realize you sent a prompt with a typo or an incomplete instruction.
A small quality-of-life fix: the queued messages list no longer shows blank rows or inflated counts, so what you see in the queue is exactly what is there.
Grok 4.5 model support
Copilot CLI v1.0.76 adds Grok 4.5 to the model picker. This is the latest model from xAI and is available alongside the existing Claude, GPT, and Gemini options. If you want to try Grok for coding tasks, select it from the model picker or use the /model command to switch.
The addition of Grok 4.5 follows the broader trend of terminal coding agents becoming model-agnostic — you are no longer locked into a single provider's models and can pick the one that works best for your specific task.
Plugin and hook controls in /plugins
The /plugins command now includes enable and disable controls for every type of extension: plugins, instructions, agents, LSP servers, and hooks. Before this change, disabling a plugin or hook required editing configuration files manually. Now you can toggle each extension on or off from within the CLI.
This is particularly useful when you are debugging an issue caused by a misbehaving plugin or hook. Instead of removing the extension entirely (and losing your configuration), you can disable it, test whether the problem goes away, and re-enable it once you have identified the cause.
Enterprise sandbox floor
For teams using Copilot CLI in corporate environments, v1.0.76 adds a sandbox floor that enterprise administrators can enforce through managed settings. The sandbox floor tightens the user's sandbox policy — it can make restrictions more restrictive, but it cannot loosen them. This means an administrator can require that all sessions run with network restrictions or filesystem limitations, and individual users cannot override those restrictions.
The /sandbox dialog now shows which settings are managed by your organization with locked fields and a clear indication of what is enforced. This makes it possible for administrators to confirm that the sandbox policy is correctly applied without having to inspect configuration files.
Other practical improvements
Several smaller changes in v1.0.76 affect daily use:
Unsent prompts stay with their session. If you typed a message but had not sent it when you switched sessions, the unsent text now stays with the session you typed it for. Before this fix, the text would follow you to the new session, which was confusing when you switched sessions to do something completely different.
web_fetch follows redirects. The web_fetch tool now handles HTTP redirects instead of failing. When a redirect points to a different origin, Copilot asks for permission and shows you where the redirect came from, so you know exactly what URL is being accessed.
Faster diff scrolling. The /diff command now scrolls and syntax-highlights large multi-file diffs faster, reducing the lag when reviewing changes across many files.
Subagent delegation improvements. Small tasks and parallel work are delegated to subagents more efficiently, which means the main session stays responsive while the agent handles background work.
Autopilot persistence after task completion. Autopilot mode now stays selected after task_complete by default, so the agent continues working on the next task without dropping back to interactive mode. If you prefer the old behavior, set stayInAutopilot to false in your configuration.
Scheduled prompt count in footer. The status footer now shows the number of active scheduled prompts, so you can see at a glance how many timed tasks are queued without checking each one individually.
Voice mode media handling. When using voice mode on macOS or Windows, Copilot now pauses any playing media before recording your voice input and resumes it afterward, so your music or podcast does not get picked up by the microphone.
How to update
If you already have Copilot CLI installed, the update is automatic — Copilot downloads new versions in the background and prompts you to restart. You can also check your current version by running copilot --version in your terminal. The Sessions sidebar requires experimental mode, which you enable with /experimental on.
For new installations, follow the official setup instructions at github.com/features/copilot. Copilot CLI requires an active GitHub Copilot subscription, which is included with Copilot Individual, Copilot Business, and Copilot Enterprise plans.
What this means for terminal coding agents
The Sessions sidebar is Copilot CLI's answer to a problem every terminal coding agent has faced: how do you manage multiple concurrent tasks without leaving the terminal? Claude Code handles this through subagents that run in the background. Codex added named sessions and thread forking in v0.146.0. Copilot CLI's approach — a visual sidebar that shows all sessions and their status — is the most IDE-like solution in a terminal tool, which makes sense given GitHub's experience building the VS Code sidebar.
The queue manager is a feature that no other major terminal coding agent offers yet. Being able to reorder, edit, and remove queued messages before they are sent gives you finer control over how the agent spends its time, which becomes important when you are batching multiple small tasks in sequence.
Together, these changes move Copilot CLI closer to the multi-session workflow that developers have been requesting since the tool launched. The single-session limitation was one of the most common reasons developers cited for sticking with IDE-based agents like Cursor, even when they preferred the terminal workflow. With the Sessions sidebar, that tradeoff gets smaller.
Top comments (0)