What Can Aiglade's Chinese Interface Actually Do? Six Tasks That Need No Command Line

Last updated: 2026-10-25

Aiglade is a Windows desktop application that puts Claude Code, Codex CLI, Hermes, and OpenClaw into a single window, with a Chinese onboarding wizard, ready to use after installation, and no programming required. It is built as an Electron desktop app and serves as the single console for these AI agents, with no web version. This article covers only six things the Chinese interface can do, and each one of them needs no command line.

What does handing off a task in Chinese look like?

The first task is handing off work in Chinese without writing a command. The user does not need to write code or explain a technical implementation. Describing the task itself in Chinese is enough, and the task then enters dispatch and queueing. The agents inside the window include Claude Code, Codex CLI, Hermes, and OpenClaw, all in the same window, so the user does not need to understand each one separately or learn each one's usage first.

The starting point is installing the app and opening it. The interface appears in Chinese, the wizard explains what comes next in Chinese, and the user follows the Chinese prompts. There is no need to learn an English interface first, and no need to master programming terms first. For a basic introduction to the product, see the Aiglade about page. For the official background of individual tools, see the Claude Code documentation and the Codex documentation. The user does not need to read those documents in full before starting.

How do you see dispatch and queue status without a terminal?

The second task is seeing dispatch and queue status in the window, without tracking progress through a command line or a configuration file. Dispatch answers the question of who the task goes to. Queueing answers the question of what order tasks run in. Together, they give several agents a predictable order, and the user can see in the interface whether a task is currently being dispatched or queued.

Queueing also keeps tasks from interfering with each other. When several tasks exist at once, the queue relationship lets each task's progress be tracked on its own, so the thread is not lost as the number of tasks grows. The user does not need to remember who a task went to, or check back and forth across different places. The task, the dispatch result, and the queue status all belong to the same management interface, and the user can map them directly to a specific task when looking. Task status itself is also readable in Chinese in the interface, including queued, waiting for confirmation, running, and marked resumable after an interruption. Status changes are readable information, not internal details that require a command line to check.

What happens on a confirmation card?

The third task is approving or rejecting on a Chinese confirmation card. When an action triggers a gate, it does not run directly. It stops and waits for the user to confirm. What the user sees is a confirmation card with two actions: approve or reject, handled through a confirmation endpoint. Approval is judged by object equality, not by keyword matching. A gate carries a timeout, and a timeout does not mean default approval.

Confirmation itself happens in Chinese, and the user can only truly decide if the content to be confirmed is readable. Confirmation covers only the single action that triggered the gate. It does not change the goal of the task, and it does not attach approval conditions to other actions. The user does not need to predict every risk before an operation starts, because the window stops at a risky operation on its own and hands the decision back to the person. This arrangement leaves judgment to the person and execution to the agents inside the window. The user takes part in decisions that are readable, not in technical details that are not.

How do you tell how far a task has gone?

The fourth task is seeing how far a task has gone, in the same window. The task entry, the task status, and the confirmation prompts all appear in one place, so facing a single window is enough to see where a task is headed and what state it is in. A single console means there is only one management entry for tasks, and the user does not need to search across several places to confirm a task's status.

The practice of Chinese documentation and Chinese interfaces is common on domestic platforms too. For example, the Alibaba Cloud Model Studio documentation and the Zhipu Open Platform documentation both provide explanations in Chinese. Seen together, the Chinese interface is not an added translation layer, but part of the onboarding process itself. It lets the user confirm at each step whether their understanding is correct, instead of guessing and continuing. When the user meets an unfamiliar prompt, the Chinese text can be read first, and then the decision to continue can be made. The wizard does not appear only at the start. It provides Chinese explanations whenever the user needs to understand the current interface, from opening the app to confirming a task, all in the same language, so attention can stay on the task instead of on language conversion.

What happens after a crash or a network cut?

The fifth task is resuming, by hand, the tasks that can resume after an interruption. After the daemon crashes or restarts, orphaned tasks are marked resumable and keep their checkpoints. Resuming is started by hand: the checkpoint is read, the task moves from resumable back to running, and the executor replays history from the checkpoint. When the user returns, the records show how far the task got, and the next step can be decided from there.

This task has a clear boundary. Only a task that carries a checkpoint and holds the resumable status can resume. It is not true that any task can resume from where it stopped. Tasks in other states cannot resume and return a 409 response instead. The search runtime does not support resuming. The discussion of resuming covers only these conditions, and the continuation starts from the checkpoint, not from the beginning of the task. The Chinese interface helps here by making it readable which steps can resume and which cannot, instead of describing the ones that cannot resume as if they could.

Where do you check what happened, without digging through terminal logs?

The sixth task is checking two record locations, without digging through terminal logs. The first is the status-change record: the daemon writes audit entries through a single path into an audit table, with action values covering creation, dispatch, heartbeat, status change, gate, resolution, and archiving. This is a record in a database, not a plain-text file on disk. The second is the session record: each CLI tool keeps its session records on the local disk, for example session files for Claude Code and session directories for OpenClaw. This second location is local.

The two locations answer different questions, and the audit table does not answer session content, while session files do not answer the full picture of status changes. To look back, the user needs to see both locations to form a fairly complete thread for a task. Looking at only one location makes it easy to mistake session content for status changes, or status changes for session content. For comparison, the Microsoft Learn documentation and the DeepSeek API documentation provide their own platforms' explanations, but this article's claims about Aiglade stay with the audit table and the local session files.

What does the Chinese interface not promise?

Beyond the six tasks, there are three things the Chinese interface does not do, or rather does not promise. The first is that not every operation requires approval. Only actions that trigger a gate stop and wait for user confirmation, and this article makes no claim about confirmation for other operations. The second is that not every task can resume. Only tasks that carry a checkpoint and hold the resumable status can resume by hand, and the search runtime does not support it. The third is that there is no web version. The Chinese interface exists only inside the Windows desktop window, and the scenario of opening a browser on another computer and picking up work is out of scope.

The three boundaries together with the six tasks give the complete picture of what the Chinese interface can do. The six tasks all need no command line, and the three things it does not do cannot be filled in with a command line either, because they are not gaps a command line could fill. They are boundaries of the product itself. Judge first whether the six tasks cover daily needs, then judge whether the three boundaries contain any wall that cannot be worked around. The six tasks share one trait: each action is a visible action on the Chinese interface. Handing off a task, seeing dispatch and queue status, tapping a confirmation card, seeing task status, resuming by hand, and checking two record locations all need no command line first, and none of them needs command-line concepts to be understood first. The test for enough is whether the six tasks include something done every day. The test for acceptable is whether the three boundaries include any wall that cannot be worked around.

A first-time user can test all six tasks in a single session without preparing anything special. Open the app and hand off a task in Chinese, then watch the dispatch and queue status in the same window. When an action triggers a gate, the confirmation card appears and the choice is approve or reject. After the task moves ahead, the status shows how far it got, and the two record locations show what changed and what each session recorded. If the session is interrupted, only a task with a checkpoint and the resumable status offers the manual resume path. Running through these six checks once gives a concrete sense of the interface, instead of a general impression from a feature list.