Install it on your own
Windows machine.

The personal edition is for one researcher, installed on your own machine, with your files and analyses staying on it. The enterprise edition targets departments, labs, hospitals and universities: self-hosted deployment, per-user isolation, institutional account authentication and deployment support. The personal edition is $50 on Gumroad; the enterprise edition is quoted by email, deployment support included. See how to get it.

The platform, the Conda environments, the raw data and the computation all sit on your own machine. What needs network access is the model call — the platform reaches the API of whichever provider you configure. If you need the inference to stay off the network too, switch the settings to a local Ollama model so the conversation content stays on the machine too.

Yes, you just risk running out of memory on large datasets. 16 GB handles ordinary projects; single-cell work is memory-hungry — cell-cell communication on 100,000 cells alone needs around 16 GB — so go for 32 GB if your datasets are large.

No. The whole Linux analysis environment is packaged in the image; import it and work straight from the browser. No second server to provision, no hand-built R/Python environment — which is exactly the first barrier described in the pain points Q&A.

No. The platform supports five model providers — Anthropic, AWS Bedrock, DeepSeek, Kimi and Ollama. Put your own API key in the settings and switch whenever you like; with Ollama the model runs locally.

No. The analysis runs inside the imported Linux environment; the browser is only the interface. Closing the tab does not affect a running job, and the progress is there when you come back. Once IM is wired up, progress cards are pushed to your phone as well.

After purchase we provide a download link via cloud storage. The archive is about 5 GB and the extraction password is www.cc-bioinfo.com. Then follow the three steps above.

Yes. The fastest way is the trial platform at test.cc-bioinfo.com (no sign-in needed) — no install, no environment setup, and no need to prepare data first. For a dedicated trial environment, email tangmoogmoogtang@gmail.com.

On the Gumroad channel the zip installer runs 9 automatic steps in a measured 4–5 minutes (up to about 15 on a slow connection); most of the time goes to downloading bun, Node.js 20, nginx and the plugin marketplaces. On the mainland-China image channel you unpack the ~5 GB archive and import it into WSL with one command — the full demonstration in video 01 is 44 seconds, and real time depends on disk speed.

The platform itself is about 1 GB (server and frontend ~312 MB, the two skills and plugin marketplaces ~686 MB); with the system dependencies the installer pulls in, actual disk growth is about 2 GB — leave at least 3 GB for the platform. The 300 GB is not platform usage: it is the recommended headroom for your raw data, per-project Conda environments and intermediate results, which single-cell and spatial studies reach easily.

Internet access is required during installation: the installer downloads bun, Node.js 20, nginx and the plugin marketplaces. Day-to-day use afterwards works offline, with one exception — calls to your model provider. Choose a local Ollama model and even that step stays on the machine.

No. The installer registers the platform to start at boot; it comes back automatically after a restart and projects resume from their last state. The personal edition has no login page — open the local address in a browser and you are in the workbench.

Neither. Supported forms are Windows 10/11 with WSL2, native Linux (Ubuntu 24.04 or newer) and virtual machines; macOS and bare Windows are outside the supported range, which the license and refund terms also state — an environment that does not meet the requirements is not grounds for a refund.

No. Routine omics analyses — differential expression, enrichment, single-cell clustering and annotation, Mendelian randomization, colocalization — run on CPU, and memory matters far more than a GPU. If your machine has one, confirm on first use that the platform detected it: in our own testing a GPU passed through to WSL2 was once misjudged as "no GPU" — see section 05 of How to choose.

Licensed per researcher: one named individual on up to 2 machines you personally use (a laptop plus a lab workstation, say). The license follows you, not the hardware, so replacing a machine is allowed. The personal edition may not be used to provide a multi-user service. Full text on the license and refund page.

Gumroad channel: unpack the new zip and rerun setup.sh; the installer detects the existing installation and upgrades in place, keeping sessions and project data — 113 sessions verified intact before release. Image channel: updated images are provided with each release; follow the notes that come with the package.

bio-design (study design) and bio-analyze (analysis execution) ship inside the package and are there as soon as you import — nothing to install separately; their versions move with each release, so the notes inside the package are authoritative. Both skills can also be installed into Claude Code on their own; what the platform adds is the isolated, resident execution environment and the complete audit trail.