Omics-OS Cloud
Managed multi-omics platform powered by Lobster AI - analyze biological data through natural conversation with 13 specialist agents
Omics-OS Cloud
Omics-OS Cloud is a managed bioinformatics platform at app.omics-os.com that lets you analyze multi-omics data through natural conversation with Lobster AI.
Get started free — Sign up at app.omics-os.com and start analyzing with unlimited prompts on the Free tier.
What You Can Do
Talk to Lobster in plain English and it will:
- Search literature — Find papers on PubMed, extract methods, discover datasets across GEO, SRA, and PRIDE
- Analyze single-cell data — QC, clustering, differential expression, cell type annotation with Scanpy and scVI
- Process bulk RNA-seq — Normalization, batch correction, pathway analysis with pyDESeq2
- Explore proteomics — Parse DDA/DIA mass spec data (MaxQuant, DIA-NN, Spectronaut, Olink), imputation, enrichment
- Run genomics workflows — VCF parsing, GWAS analysis, variant annotation
- Perform machine learning — Feature selection, survival analysis, biomarker discovery, multi-omics integration
- Visualize results — Publication-quality UMAP, heatmaps, volcano plots, violin plots via Plotly
How It Works
- You ask — Describe your analysis in natural language
- Lobster delegates — The supervisor routes your request to specialist agents
- Agents execute — Domain experts run validated bioinformatics tools on your data
- Results stream — See results, plots, and insights in real-time via SSE
You: Find scRNA datasets on senescence in human fibroblasts,
download top 3, run QC, cluster cells, identify cell types
[Supervisor -> Research Agent]
Searching GEO for scRNA-seq datasets on senescence...
Found 5 matching datasets.
[Supervisor -> Data Expert]
Downloading GSE167829, GSE182118, GSE201910...
Converting to H5AD format.
[Supervisor -> Transcriptomics Expert]
Running QC: 24,531 cells pass filters.
Clustering: 8 distinct populations identified.
Cell types annotated: Fibroblasts, Immune, Endothelial.
[Supervisor -> Visualization Expert]
Generating UMAP plot and marker gene heatmap.
Done in 12 minutes. Jupyter notebook exported.Modular Agent System
Omics-OS Cloud runs on Lobster AI v1.0.0 — a modular multi-agent architecture with 13 specialist agents organized into 8 installable packages.
Supervisor-Accessible Agents (9)
| Agent | Package | What It Does |
|---|---|---|
| Transcriptomics Expert | lobster-transcriptomics | Single-cell and bulk RNA-seq analysis (delegates to sub-agents) |
| Research Agent | lobster-research | Literature search, dataset discovery, GEO/SRA queries |
| Data Expert | lobster-research | Offline data download and format conversion |
| Proteomics Expert | lobster-proteomics | Mass spec (DDA/DIA), affinity proteomics, protein quantification |
| Genomics Expert | lobster-genomics | VCF parsing, SNP arrays, GWAS analysis |
| ML Expert | lobster-ml | Feature selection, survival analysis, biomarker discovery (delegates to sub-agents) |
| Visualization Expert | lobster-visualization | Publication-quality figures via Plotly |
| Metadata Assistant | lobster-metadata | Cross-database sample annotation and harmonization |
| Structural Viz Expert | lobster-structural-viz | Protein structure visualization (PyMOL, ChimeraX) |
Sub-Agents (4)
These agents are not directly accessible — they are delegated to by their parent agents for specialized tasks:
| Agent | Parent | What It Does |
|---|---|---|
| Annotation Expert | Transcriptomics Expert | Cell type annotation and QC |
| DE Analysis Expert | Transcriptomics Expert | Differential expression analysis |
| Feature Selection Expert | ML Expert | Feature ranking and selection |
| Survival Analysis Expert | ML Expert | Clinical outcome prediction |
Architecture
Omics-OS Cloud runs on AWS with four independent CDK stacks:
| Component | Technology | Purpose |
|---|---|---|
| Backend | ECS Fargate (ARM64, 16 vCPU, 120GB RAM) | FastAPI processing with Lobster AI agents |
| Frontend | S3 + CloudFront + WAF | React UI with assistant-ui framework |
| License Service | Lambda + API Gateway + DynamoDB | Stripe billing, Cognito tier management |
| Auth | AWS Cognito | JWT-based authentication with tier claims |
Streaming: HTTP SSE with DataStreamResponse protocol (not WebSocket). Real-time agent output as it executes.
LLM: Google Gemini 3 Pro (production default). BYOK (Bring Your Own Key) supported for OpenAI, Anthropic, and other providers.
Provenance: Every analysis step is logged with W3C-PROV compliant audit trails, exportable as Jupyter notebooks.
Get Started
Quick Start
Your first analysis in under 5 minutes
API Reference
REST API and streaming protocol documentation
Agent Packages
Explore the 8 modular agent packages
Privacy & Security
How your data is protected
Pricing
| Tier | Price | Compute Limits | Agents |
|---|---|---|---|
| Free | Free (signup required) | 60s timeout, 50K cells, 10 tool calls/request | All free-tier agents |
| Cloud | $49/month per user | 5 min timeout, 500K cells, 50 tool calls/request | All agents including premium |
| Enterprise | Custom | Unlimited timeout, unlimited cells, unlimited tool calls | All agents + custom development |
For Developers
Omics-OS Cloud exposes a REST API for programmatic access:
- Streaming chat:
POST /sessions/\{id\}/chat/stream(SSE with DataStreamResponse format) - Session management:
GET/POST/DELETE /sessions - File upload:
POST /files/upload(H5AD, CSV, VCF, 10X formats) - Modalities:
GET /modalities(list loaded datasets) - Plots:
GET /plots(retrieve generated visualizations)
All endpoints require JWT authentication via AWS Cognito.