Qwen Code,
in your notch.

Qwen Code notifications, permission approvals and cost on your Mac

Gnista shows every Qwen Code session in your MacBook’s notch: working, done, or waiting for you. Allow or deny its commands there, reply when a turn is done, and see what each session costs.

Try free for 7 days No account, no card · then €19 once

01At a glance

Gnista with Qwen Code, at a glance.

Where it runs: Qwen Code in any terminal, with any model provider Qwen supports.
Setup: Hooks in ~/.qwen/settings.json, after a backup.
Needs you: Permission requests with the command or file, and Qwen’s questions and plans.
Answer from the notch: Allow or Deny for permission requests; questions and plans are answered in Qwen.
Reply from the notch: Yes, typed into Qwen’s terminal: Terminal, iTerm, Ghostty, a tmux pane, or an editor’s terminal with Gnista’s extension.
In the row: The session’s title, your latest prompt and Qwen’s reply, the tool in use, todo progress, context and cost.
Subagents: Shown in the row while they run, background ones included.
Cost and limits: Tokens and cost by model from Qwen’s session file; no plan limits, as Qwen records none.

Allow. Deny. Keep going.

When Qwen Code asks for permission, the notch opens with the command or file. Allow it and Qwen runs it; deny it and Qwen reads “Denied from Gnista”. Leave it, and Qwen asks in its own prompt: at once if you go to its tab, after 10 seconds with the notch closed, 45 at most while you’re in it.

Questions and plans, without the wait.

Qwen asks its questions and shows its plans only in its own prompt. Gnista doesn’t hold them: the row says Qwen has a question or a plan is ready, and Qwen asks you straight away. One click lands in its tab.

The whole session, in one row.

Gnista reads Qwen’s session file: the title you gave it with /rename, your latest prompt and its reply, the tool it’s running, todo progress, the context ring, and subagents working in the background.

What it costs, by model.

Each reply’s tokens are priced at its model’s rate, on Alibaba’s Token Plan or any other provider. For a model Gnista doesn’t know, such as a local one, describe it in a model file to give it a price and a context size.

What it can’t do with Qwen Code, and why.

Qwen’s questions and plans can’t be answered from the notch: Qwen takes no hook’s answer for them. The row says Qwen has a question or a plan is ready, and Qwen asks in its own prompt at once.
There’s no Always: Allow and Deny answer one request at a time.
Qwen’s “potential loop” warning runs no hook, so it doesn’t show in the notch.
Replies from the notch are typed into Qwen’s terminal, so they work only in Terminal, iTerm, Ghostty, a tmux pane, or an editor’s terminal with Gnista’s extension.
Gnista writes to ~/.qwen/settings.json. If you set QWEN_HOME, Qwen reads its settings from there and doesn’t see Gnista’s hooks.

02Setup

Connected in a minute.

Open Gnista and tick Qwen Code.
Gnista backs up ~/.qwen/settings.json to ~/Library/Application Support/Gnista/Backups and adds its hooks, keeping yours.
Start a Qwen Code session. Nothing to switch on in Qwen: its hooks are on by default.

03FAQ

Qwen Code, asked and answered.

Can I approve Qwen Code from the notch?

Yes: Allow or Deny. If you leave it, Qwen asks in its own prompt: at once when you go to its tab, after 10 seconds while the notch is closed, after 45 seconds at most while you’re in the notch.

Can I answer Qwen’s questions and plans from the notch?

No. Qwen ignores a hook’s answer for them, so the row tells you, and Qwen asks in its own prompt at once.

Does it show what Qwen Code costs?

Yes: tokens and cost by model, read from Qwen’s session file. Models without a public price, such as local ones, need a model file.

Can I reply to Qwen Code from the notch?

Yes, in Terminal, iTerm, Ghostty, a tmux pane, or an editor’s terminal with Gnista’s extension: Gnista types your reply into Qwen’s terminal.

Does it work with my model provider?

Yes. Gnista reads Qwen’s hooks and session file, not the model, so any provider Qwen supports works: Alibaba’s Token Plan or a local OpenAI-compatible server, for example.