How to Try LynxKite Without Committing to a Program

You can try LynxKite in three ways, none of which require signing up for a commercial program: run the live interactive workspace demo on lynxkite.com, watch the recorded demo, or use the open-source version from GitHub. The live canvas is the fastest way to see the pipeline mechanics — you click boxes to inspect them and run the pipeline on GPU — while GitHub is the right entry point if you want to evaluate the code itself.

Option 1: The live interactive workspace

The homepage hosts a live canvas, not a video. According to the site, you can click any box to inspect it, then hit Run on GPU and watch the orchestrator execute the pipeline stage by stage, with GPU microservices starting and stopping automatically.

The demo pipeline shown is a compound-screening workflow with these stages:

A useful detail from the canvas: GPU nodes are marked as up and auto-scaled, and the site states that open-model GPU microservices start and stop with the workflow, so idle GPU time does not land on a program budget. That is the mechanism worth watching for if you are evaluating cost behavior rather than just features.

What to do: open lynxkite.com, click through the boxes in the canvas to see each stage's configuration, then run it and watch the stage-by-stage execution. If you want to see it built rather than just run, the "Watch Demo" link covers the pipeline being assembled live.

Option 2: Watch the demo

The site offers both Try the demo and Watch Demo entry points. Use the recorded demo if you want the narrated walkthrough first — it is the lower-effort way to decide whether the workflow model fits your problem before you spend time clicking through the live canvas.

Option 3: Open source via GitHub

LynxKite is described as open core: free and open source, with a commercial tier for production teams. The GitHub link is in the site navigation. This is the path to take if you want to inspect the implementation, check the 600+ graph operations, or run it against your own data rather than the demo graph.

What you are actually evaluating

The site positions LynxKite as agentic, multi-scale modeling: LLM agents orchestrate open and proprietary models from genes to molecules, with RAPIDS/cuGraph under the hood for graphs that don't fit on a laptop. The stated pipeline is:

1. Identify the target with graph reasoning
2. Screen compounds (claimed 50,000× faster molecular dynamics trajectory prediction)
3. Predict toxicity
4. Benchmark against your protein of interest

It also supports running any Hugging Face model as a LynxKite Inference Microservice — a plain Python row function becomes a scaling GPU service on Kubernetes via a single @lim decorator, per the site.

For production context, the site states LynxKite powers Biophytis' AI-driven longevity drug-discovery platform (the MASSIVE project), showcased with Biophytis, Nebius, and Accenture at NVIDIA GTC 2026. LynxKite is also listed as an NVIDIA Inception Premier member built on NVIDIA RAPIDS & CUDA.

Where to start reading

• Documentation — linked in the main navigation, the place to go for setup and operation details.
• FAQ — linked in the navigation, useful for quick answers before you commit time to the demo.
• Use Cases and Life Sciences / LIMs pages — for judging fit against your own workflow.

Practical notes

• The site does not publish pricing on the pages referenced here, so treat the commercial tier as "contact required" rather than assuming any free tier beyond the open-source version.
• The demo runs on GPU infrastructure the site controls; running the open-source version yourself means providing your own GPU environment.
• If your goal is a quick sanity check on whether graph reasoning plus molecule analysis in one workflow matches your problem, the live canvas plus the recorded demo will answer that faster than reading docs.