GitHub Copilot CLI in NeuroSquadGitHub Copilot CLI,a whole squad of it
Every card runs the real copilot in its own terminal — on your Copilot subscription, or on your own key with no GitHub login at all. Copilot’s hooks tell NeuroSquad when each one works, asks or finishes, and arrows give it terminals, browsers, notes and other agents, live.
Official sitegithub.com/features/copilot/cli- The real
copilot, your own login or key - Nothing written to
~/.copilot - Free, for Windows 10 and 11
It knows when Copilot is done — and when it’s asking
Copilot CLI has lifecycle hooks of its own. Each card loads a small plugin whose hooks post straight to NeuroSquad, so every state comes from Copilot itself.
userPromptSubmittedhookYou press Enter and Copilot fires userPromptSubmitted. The card’s icon spins, the Squad status card counts it as working, and the toast from its last question is taken down.
notificationhookCopilot’s notification hook says it is asking — “Run command: rm -rf dist”, or an ask_user question. The card pulses and the toast leads with Copilot’s words until you answer.
agentStophookagentStop ends the turn — the main agent’s, not a subagent’s. The next queued prompt goes out, the journal gets the answer, and a toast says it’s done.
Only real questions
The status follows Copilot’s notification, not its permissionRequest, which fires even when --yolo would approve the call anyway.
Cancelled turns too
Ctrl+C fires no hook, so when the idle footer returns mid-turn NeuroSquad checks Copilot’s event log for abort — and the card finishes.
Click, and you’re there
A toast click switches to the workspace, flies the camera to the card and flashes it — even among dozens of agents.
Three things you’ll do every day
Replicas of the app, played step by step. When Copilot asks for permission, the answer is yours.
Copilot fixes a failing test, asks before deleting a folder, and finishes — the journal writes itself.
Everything GitHub Copilot CLI can do in NeuroSquad
Copilot CLI has hooks, plugins, live tool lists and a telemetry file — enough for everything NeuroSquad does for Claude Code. Here is all of it, with a docs page for each.
- 01
Knows its state
Copilot’s own hooks report every turn, so nothing is guessed.
Finished and waiting are different
A plugin per card subscribes to Copilot’s hooks:
Docs: Finished and waiting are differentuserPromptSubmittedmarks it working,notification— waiting on you,agentStop— finished.A notification that says what it needs
The toast leads with Copilot’s words —
Docs: A notification that says what it needsRun command: rm -rf dist, or the question it asks you. One per agent; a waiting one stays until you answer; a click flies to the card.Every card is its own session
Docs: Every card is its own session--session-idwith the card’s id both creates the session and resumes it, so a restart continues the conversation./clear,/newand/resumeare followed through the hooks.The Squad status card
Every agent of the workspace with its status, the model of its latest call and the live time of its turn.
Docs: The Squad status cardNo surprises at start
The folder-trust question is answered “Yes” for this session only, as you trusted the folder by opening it. And Copilot’s offer to edit Windows Terminal’s own settings is switched off with
Docs: No surprises at startCOPILOT_SETUP_TERMINAL=false.
- 02
Works through the canvas
An arrow from the card is a set of tools — and Copilot picks them up mid-session, no restart.
Arrows are tools, live
An arrow to a browser, terminal, note or another agent gives Copilot
Docs: Arrows are tools, liveneurosquad-browser_*,terminal_*,note_*oragent_*. Copilot handlestools/list_changed, so they arrive in its very next request.It can add cards itself
Docs: It can add cards itselfcanvas_spawn_cardlets it open a terminal, a browser, a note or a helper agent next to itself, already joined by an arrow.Canvas mode
Copilot starts with its PowerShell, bash and web tools excluded and
Docs: Canvas mode--deny-tool=shell. Copilot’s denials beat--yolo, so commands always run in terminal cards you can watch.MCP servers from the registry
Install a server from the official MCP Registry once, then give it to an agent with an arrow. Copilot asks before every call to it, and its secrets never leave NeuroSquad.
Docs: MCP servers from the registrySkills from skills.sh
An arrow to a skill card gives Copilot
Docs: Skills from skills.shskill_loadwith each skill and when to use it — live, like every other tool.
- 03
Built for long runs
Full contexts and rate limits, both handled on the card.
Context meter and Compact
The ring in the header shows how full the window Copilot compacts against is, from its own telemetry. Compact sends
Docs: Context meter and Compact/compact.Continues after a rate limit
When Copilot hits a rate limit, the card reads the reset from its message (“Resets in 2h 15m”), counts down and sends “continue” then. Off per card if you prefer.
Docs: Continues after a rate limitPrompt queue
Line up the next messages: each goes out on
Docs: Prompt queueagentStop, never on a pause mid-thought. Daily prompts too.Hand off to a fresh agent
When a session is full, start a new agent beside it — Copilot or any other CLI — with a summary and Copilot’s last answer typed in, ready for you to send.
Docs: Hand off to a fresh agentJournal
Each finished answer, read from Copilot’s session log, is added to a connected note: a running log of what the agent did.
Docs: Journal
- 04
Works in a team
Several Copilot agents — or a mix of CLIs — on one project without stepping on each other.
Isolated worktrees
Give an agent its own git worktree, so parallel agents never edit the same files. A setup command and files to copy, like
Docs: Isolated worktrees.env, prepare it first.Add a reviewer
One click starts an agent on a different CLI in the same worktree, with an arrow from the author and a prompt to review without editing anything.
Docs: Add a reviewerDangerous mode, per card
For a run you trust, a card starts Copilot with
Docs: Dangerous mode, per card--yolo— all tools, paths and URLs allowed. Off by default; canvas mode still holds.Roles
Reviewer, tester, researcher, architect, writer or your own: the role is typed in as a first prompt you can see and edit.
Docs: RolesOne instructions file
Copilot reads AGENTS.md and
Docs: One instructions file.github/copilot-instructions.mditself, so the file your other agents use works for Copilot as is.
- 05
Cost, control and access
What it used, what it may spend, which model it runs on, and how you reach it.
Usage per request
Copilot keeps no per-request tokens in its logs, so each card points its OpenTelemetry file exporter at a local file: one record per model call, subagents included. Nothing leaves your machine.
Docs: Usage per requestBudget brake
A limit per workspace. Past it, each agent gets one Ctrl+C — Copilot’s turn stops, the session stays — and automatic prompts stop until you raise the limit.
Docs: Budget brakeOpenRouter, no GitHub login
A card can run on an OpenRouter model through Copilot’s own bring-your-own-key settings — no Copilot subscription needed. The key goes only into that card’s environment.
Docs: OpenRouter, no GitHub loginDictation into the prompt
Hold a hotkey and speak. Speech is recognised on your machine and typed into the active agent’s prompt.
Docs: Dictation into the promptFrom your phone
Scan a QR code and the whole canvas opens in the phone’s browser: read Copilot’s terminal, answer its question, send the next prompt.
Docs: From your phone
Who’s waiting, and what it used
Two cards that answer the questions you ask most when several Copilot sessions run at once.
The Squad status card
Every agent of the workspace with its status, model and turn time. The one waiting on you comes first, with what it asks. Tap a status to filter.
- coupon-fixGitHub Copilot CLIgpt-5.4GitHub Copilot needs your permission: Run command: rm -rf distNeeds you2:34
- api-migrationGitHub Copilot CLIclaude-sonnet-4.5Move routes to the new clientWorking11:07
- search-uiGitHub Copilot CLIopenrouter/autoRework the search boxWorking4:08
- docs-passGitHub Copilot CLIgpt-5.4Update the READMEFinished7 min. ago
Usage & cost
One record per model call from each card’s own telemetry file. Models Copilot serves itself bill in AI credits, not tokens, so they have no price here — never a made-up $0.
| Agent | Tokens | Cost | ||
|---|---|---|---|---|
| coupon-fix | 1,046,220 | no price | ||
| api-migration | 713,408 | no price | ||
| docs-pass | 198,312 | $0.34 | ||
| Total | 77 | 1,957,940 | $0.34 |
The real Copilot CLI, plus a few arguments
NeuroSquad starts the Copilot CLI you installed — its own copilot.exe — in a real terminal in the card. This is everything it adds:
Nothing written to ~/.copilot
The plugin, the MCP config and the telemetry file are per card in NeuroSquad’s own data folder.
No secret on disk
The per-card token lives only in the environment; Copilot expands it into the MCP and hook headers itself.
Your login, or your key
Copilot runs on your GitHub Copilot subscription — or, on OpenRouter, on your own key with no GitHub login at all.
GitHub Copilot CLI in NeuroSquad, answered
Give Copilot a canvas
Free for Windows 10 and 11. Your code, your login and your keys stay on your machine.