MaaNTE: a MaaFramework assistant for Neverness to Everness
MaaNTE. Nevertheless to Everless automatic assistant 异环小助手
At a glance
- What is it?
- MaaNTE automates fishing, coffee-making, furniture collection and other daily loops in Neverness to Everness by reading the game window with computer vision. It is a Windows-only, AGPL-3.0 Python project built on MaaFramework, and its own README warns that using it may violate the game's fair play policy.
- Who is it for?
- MaaNTE is for players on Windows who accept the account risk and want the fishing, coffee, furniture and reward loops handled for them; the README points ordinary users at the GitHub Releases page rather than a source clone. Anyone unwilling to run a third-party input simulator against a game whose own fair play statement lists automatic idling as a violation should not install it at all.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository last received commits 10 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 17, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What MaaNTE automates in Neverness to Everness
MaaNTE is a third-party assistant for the game Neverness to Everness (异环). The README describes it as a tool that recognises the game screen and simulates ordinary interaction to simplify repetitive play. It does not modify game files or game data, according to its own disclaimer, and it is free and open source.
The feature list is a set of daily chores rather than a combat bot. It covers automatic fishing with automatic fish selling and bait buying, automatic coffee making with customer clearing, collecting income from the coffee shop with restocking and product rotation based on an in-game wind vane, collecting furniture across five apartments (维纳, 伊登, 天景, 金都 and 峰林), and claiming activity and 环期赏令 rewards. There is also a 粉爪大劫案 mode with an infinite loop option plus earnings statistics and logging, an automatic Tetris mode with a built-in AI that repeats until stamina runs out, and a real-time assistance group covering auto-teleport, story skipping and auto-pickup.
Two features are unusual for this class of tool. Auto-dodge is described as based on audio recognition rather than only screen capture, and the music features include looping a specific song (迷星叫) and playing arbitrary MIDI files on the in-game piano. Preset task groups split the work into an idle group and a real-time assistance group, which matters because the two modes have different constraints.
It is aimed at players who already run the game on Windows and want the routine loops removed. It is not a general-purpose automation framework, and nothing in the README suggests it works on any platform other than Windows.
How MaaFramework drives the automation
MaaNTE is built on MaaFramework, which the acknowledgements section describes as an image-recognition-based black-box testing framework. That choice explains most of the project's shape. MaaNTE does not read game memory and does not inject into the game process. It captures the window, matches what it sees against templates and pipeline definitions, and then sends input events.
The repository layout reflects this split. The top level holds a Python requirements file, a package.json used for MaaFramework tooling (@nekosu/maa-tools), a maatools.config.mts configuration file, and directories named agent, assets, deps, docs, scripts and tools. The Python dependencies in requirements.txt show what the runtime actually needs: maafw pinned to v5.10.4, OpenCV, onnxruntime-directml, librosa, numpy, scipy, scikit-learn, soundcard, mido, scapy, pktmon-interface and websockets. The presence of onnxruntime-directml indicates inference runs locally on DirectML rather than through a cloud service. librosa and soundcard line up with the audio-based dodge feature. mido lines up with MIDI playback for the piano feature. scapy and pktmon-interface suggest some network-level capture, though the README does not explain what that path does.
The practical consequence of a vision pipeline is fragility. Anything that changes the game's UI, resolution or rendering can break template matching, which is why the README pins the game to 1280x720 in windowed mode. The fishing algorithm is noted as supporting 120 FPS, which implies frame timing matters for at least that task. Because the pipeline is defined in data files rather than compiled code, fixes for a moved button are usually asset and pipeline edits, not Python changes.
Installing MaaNTE and running a first task group
The README is explicit that ordinary users should not clone the repository. It says to download the Release build from the GitHub Releases page, and to clone only if you have development needs. There is also a MirrorChyan mirror linked in the badges, and the README warns that copies obtained outside official channels may contain malware and are generally not the newest version.
The official documentation site, docs.maante.org, is where download and troubleshooting guidance lives. The README's own troubleshooting section points to a manual at docs/zh_cn/问题排查.md before anything else. If you do clone the repository for development, the Python dependencies are pinned in requirements.txt:
pip install -r requirements.txtThat installs maafw==v5.10.4 along with OpenCV, onnxruntime-directml, librosa, soundcard, mido and the rest of the runtime stack. The JavaScript side is managed with pnpm, and package.json lists @nekosu/maa-tools as a dev dependency, so pipeline tooling is installed separately from the Python runtime.
Before starting any task, the game must be running in windowed mode at 1280x720. The README states this as a requirement, not a recommendation, and the fishing algorithm additionally supports 120 FPS. Once the game is at that resolution, the workflow the README describes is to pick a preset task group: the idle group (挂机任务) for unattended loops such as fishing, coffee and furniture, or the real-time assistance group (实时辅助) for teleport, story skip and pickup while you are actually playing. The README does not document rollback or an undo path for in-game actions the assistant performs, so treat the first run as something to watch rather than leave unattended.
The account risk MaaNTE's own README states
The most important limitation is not technical. The README quotes the game's fair play statement (《异环》公平游戏宣言), which prohibits third-party tools that break fairness and names automatic idling (自动挂机) among the behaviours it will act against. The stated penalties include deducting illegitimate gains, freezing the account, and permanent bans. The README's own caution block tells the user to fully understand and voluntarily accept all risks of using the tool.
That is a real constraint, not boilerplate. No amount of image recognition or avoidance of memory modification changes the fact that the game operator has publicly listed automatic idling as a violation. The disclaimer also notes that the project is for learning and exchange only and must not be used commercially.
Beyond policy, there are practical failure modes. A vision pipeline depends on the screen looking the way the templates expect, so a UI update, a different resolution, or a window that is obscured or backgrounded can cause a task to misfire. The README's troubleshooting manual exists precisely because things go wrong. The project also describes itself as still in development and welcomes pull requests and issues, so behaviour can change between releases; v1.4.0, v1.4.1 and v1.4.2 all landed within about a week of each other in late August 2026.
If you want an assistant that carries no account risk, this is the wrong tool, and there is no configuration in the README that removes that risk.
MaaNTE compared with OK-NTE and the wider Maa family
The acknowledgements section names a direct alternative: OK-NTE, described as an image-recognition automation tool for the same game that supports background operation. That is the concrete difference. Running in the background means the game window does not need to be the foreground window for the automation to proceed, which matters if you want to use the machine for something else while tasks run. MaaNTE's README pins the game to a 1280x720 windowed mode and does not claim background operation, so the two tools make different trade-offs between visibility and usability.
The rest of the acknowledgements are siblings rather than competitors: MaaAssistantArknights for Arknights, MaaEnd for Arknights: Endfield, and M9A for Reverse: 1999. All of them sit on MaaFramework, so if you have used any of those, the pipeline model, the MXU interface (MaaFramework Next UI, also credited) and the general shape of the task configuration will be familiar. Choosing between MaaNTE and another Maa-based assistant is mostly a question of which game you play, not which framework is better.
One entry in the acknowledgements is struck through: MFAAvalonia, the Avalonia-based generic GUI for MaaFramework. The strikethrough implies it is no longer part of the stack, which is consistent with MXU being the credited UI instead. If you are following an older guide that tells you to install MFAAvalonia for MaaNTE, that guide is out of date relative to the README.
Licence, maintenance and what upgrades cost
MaaNTE is licensed under GNU Affero General Public License v3.0 only, identified in the README by its SPDX identifier AGPL-3.0-only. The practical implication of the Affero variant is that if you modify the software and let users interact with it over a network, the licence requires you to offer them the corresponding source. That is a stronger obligation than the plain GPL, and it is worth reading the licence text itself rather than relying on a summary. Nothing here is legal advice.
The repository is not archived. The last push was on 2026-09-08, and releases v1.4.0, v1.4.1 and v1.4.2 were published on 2026-08-22, 2026-08-29 and 2026-08-30 respectively. That pattern of frequent patch releases suggests fixes ship quickly, and it also means upgrade cost is not zero: a patch can change pipeline assets, and a task that worked yesterday can behave differently today. There is no long-term support branch mentioned in the README, and the default branch is dev, which is a development branch rather than a stabilised release line.
For a user, the upgrade path the README endorses is downloading a new Release build rather than pulling source. For a contributor, the cost is higher: the project spans Python and a separate pnpm-managed tooling layer, plus pipeline assets, so a change to one task may touch templates, pipeline definitions and Python code together. The README points developers at docs/zh_cn/develop/README.md as the entry point, which is where the actual contribution workflow is documented.
Editorial conclusion
MaaNTE is for players on Windows who accept the account risk and want the fishing, coffee, furniture and reward loops handled for them; the README points ordinary users at the GitHub Releases page rather than a source clone. Anyone unwilling to run a third-party input simulator against a game whose own fair play statement lists automatic idling as a violation should not install it at all. Before running it, confirm the game is at 1280x720 windowed, check the problem-troubleshooting manual at docs/zh_cn/问题排查.md, and read the AGPL-3.0-only terms if you plan to redistribute anything built on the code.
Frequently asked questions
Where do I download MaaNTE?
The README directs ordinary users to the GitHub Releases page rather than cloning the repository, and it warns that copies from unofficial channels may contain malware and are generally not the latest version. A MirrorChyan mirror is also linked from the project badges, and docs.maante.org is the official documentation site.
What resolution does the game need to run at for MaaNTE?
The README states the game must run at 1280x720 in windowed mode, and notes that the new fishing algorithm supports 120 FPS. This is a requirement of the vision pipeline, not a preference.
Which tasks can MaaNTE run automatically?
The feature list covers automatic fishing with selling and bait buying, automatic coffee making, collecting coffee shop income with restocking and product rotation, collecting furniture across five apartments, claiming activity and 环期赏令 rewards, 粉爪大劫案 with an infinite loop option, automatic Tetris, and real-time assistance such as auto-teleport, story skipping and auto-pickup.
Does MaaNTE work in the background while I use my computer?
The README does not claim background operation for MaaNTE; it pins the game to a 1280x720 windowed mode. The acknowledgements section names OK-NTE, a tool for the same game, as supporting background operation, which is a different approach.
What licence is MaaNTE released under?
The README states the project uses GNU Affero General Public License v3.0 only, with the SPDX identifier AGPL-3.0-only. The Affero variant adds source-availability obligations when modified software is offered to users over a network.
Community notes