InkSight: a calm e-ink desk companion with its whole toolchain in a browser

One website instead of a kit of scattered tools

InkSight is described as a calm e-ink desk companion with one website for flashing, configuration, preview, and discovering new modes, and the official site is inksight.site.

The design goal is stated as turning a small e-ink screen into a quiet, always-visible information surface for your desk. The argument is explicitly a contrast with the usual alternative: instead of another glowing notification feed, it gives useful content in a paper-like form that stays visible without adding screen fatigue.

What is interesting is that the project's main deliverable is not the hardware but the website. The README says that even if you are completely new to e-paper devices, ESP32 boards or WebSerial, the product is designed so you can get started by following the user interface step by step instead of assembling your own toolchain first. The site covers flashing firmware in the browser with the Web Flasher, configuring devices online with modes, preferences, refresh strategy and per-mode overrides, previewing content before saving so the final e-ink result is visible in advance, a no-device demo flow for trying it without hardware, and the mode plaza for discovering what other people have built.

The project also positions itself as open and extensible: firmware, backend, web configuration and the JSON mode system are all designed to be expanded over time, including future hardware design files.

The mode system and the twenty four built-ins

The content system is where most of the product lives, and the README describes it in terms of modes rather than screens.

There are 24 built-in modes, and the list is a mixture of the practical and the atmospheric. Daily Picks offers quotes, books, facts and seasonal context. The Weather Dashboard gives live weather with practical advice. Poetry, Zen and Stoic provide calm, reflective content for a focused desk. The AI Briefing covers technology highlights and AI insights. ArtWall is described as black-and-white AI artwork tailored to context. Memo, Countdown, Habit and Fitness are the everyday desk utilities.

What makes the mode system more than a template gallery is the lifecycle around it. You can create custom modes, save them to your device, share them to the mode plaza, and install community-created modes. So a mode is a piece of content definition that travels: from the person who wrote it, to the plaza, to somebody else's desk.

Because the modes are JSON, they are also reviewable, diffable and editable in a text editor, which is a genuine advantage over a configuration screen inside a firmware image. The cost is that a mode is a document with a schema rather than a stable plugin contract, so the documentation on configuration and the API pages becomes the reference that matters.

The hardware the project is designed around

The recommended hardware table is narrow, and that narrowness is a feature for anyone deciding whether to build one.

The microcontroller is an ESP32-C3 development board, the display is a 4.2 inch SPI e-paper panel, power is USB for development with an optional lithium battery build using a 505060-2000mAh cell and a TP5000 charger board, and the typical cost of a do-it-yourself build is around CNY 220.

The README is explicit that the public documentation and setup flow are centred on ESP32-C3 with 4.2 inch e-paper, and that for a first build you should start with that combination. The hardware purchasing guide at docs/en/bom.md carries the detailed parts list and part links, with a note that the links are mostly for Taobao and that equivalent parts can be found on AliExpress or Amazon.

The battery option is worth understanding before you commit, because it changes the assembly rather than just the parts list. A desk companion that stays visible all day is only useful if it is not tethered, and the TP5000 charging board plus the 505060 cell is what the recommended build uses for that.

The documentation is split by task, which is what a builder actually needs: a hardware guide, an assembly guide, a flashing guide and a configuration guide, plus a step-by-step assembly video.

Community hardware: printed cases and custom boards

The README has a community gallery section, and the entries are the kind of thing that tells you how extensible the design is.

On the mechanical side there are 3D printable cases on MakerWorld, including an Orange desktop case and a set of Pink and Red minimalist cases, and a Huawei Nova 14 colour display phone case made for a 3.98 inch four-colour e-ink screen. The phone case is the interesting one, because it shows the concept working on a device nobody designed for it.

On the electronic side there are custom printed circuit boards hosted on OSHWHUB, including an InkSight 4.2 inch custom driver board and a Moyu AI e-ink screen marked as verified.

What this tells you about the project is that the hardware reference design is a starting point rather than a fixed product. A custom driver board means people have gone past the reference wiring, and a phone case means the form factor is not locked to a rectangle on a stand.

The practical caution is the usual one for hobby hardware: none of these are supported designs, so a builder is choosing to do the integration work themselves, with the project's documentation as the starting point rather than as a guarantee.

Self-hosting the backend and the extension points

The project separates the hosted experience from the parts you can run yourself, and the documentation for the self-hosted path is specific.

There is a deployment guide at docs/en/deploy.md with a Chinese version at docs/deploy.md, an architecture document at docs/en/architecture.md, an API document at docs/en/api.md, and a plugin and extension development guide at docs/en/plugin-dev.md. The README frames the audience for these as developers who want their own local deployment, or who want to go beyond the hosted website and build custom integrations or workflows.

The repository layout matches that split. There is a backend/ directory, a firmware/ directory, a webapp/ directory for the browser side, an inksight-mobile/ directory, an openclaw_skill/ directory, scripts/, and docs/. There is also a prebuilt file named inksight-v1.0.3mf, a structure.png diagram, a CLAUDE.md, a CONTRIBUTING.md and a Chinese README.

So there are three layers a developer can work in. The firmware is where the display behaviour lives, the backend is where the content comes from and the configuration is stored, and the web app is where the user interacts. The plugin and extension guide suggests the fourth layer, where a mode or an integration is added without touching the three above.

Where InkSight is the wrong tool

Three cases point elsewhere.

The first is a display that needs to be fast or colourful. E-ink is a technology with a refresh behaviour and a colour depth that are deliberate properties rather than limitations to work around, and the mode system in this project is built around content that changes on the order of minutes or hours. A dashboard that has to update every second, or anything that needs full colour, is a different device.

The second is a person who does not want to build hardware. The project assumes a board, a panel and, for the battery version, a soldering iron. The no-device demo exists to show the experience, and the web flasher removes the need for a local flashing toolchain, but neither of those removes the need for the device. If you want a desk display to arrive working, this is the wrong project.

The third is a production product with a support commitment. The mode plaza is community content, the reference hardware is a recommendation rather than a product, and the version numbers are small, with the latest tagged release at 0.6.0. The project describes itself as open and extensible, which is the right thing for a hobby platform and the wrong thing for procurement.

One more point about maintenance. The last push to the repository was on 2026-07-15, so the project is being worked on, but a device that has to be assembled and flashed is only as good as the firmware you can find for it, and the releases are infrequent enough that a self hosted deployment should be built to survive a few months without an update.

Editorial conclusion

Adopt InkSight if you want a desk display that shows a small amount of information without glowing, if you are comfortable with a small ESP32 board and a 4.2 inch e-paper panel, and if you are willing to write or adapt JSON mode definitions. Do not adopt it if you need a general purpose display, a colour screen, or a device you can buy off the shelf and flash without a soldering iron. Verify three things first: the hardware combination, because the public documentation is centred on ESP32-C3 with 4.2 inch e-paper and the parts guide is where the panel choice matters; the mode system, since the built-in modes and the community ones are JSON documents you can edit rather than a plugin interface with a stable contract; and the hosting story, because the website is the product and self-hosting it means running the backend yourself. The licence is MIT, the last push was on 2026-07-15, and the newest tagged release is v0.6.0 dated 2026-06-20.

Frequently asked questions

Official sources

1. datascale-ai/inksight on GitHub
2. License: MIT
3. Project website
4. README
5. Releases