lobster-ai 1.1.2

Multi-Agent Bioinformatics Analysis System powered by LangGraph

Open-source multi-agent bioinformatics engine. Describe your analysis in natural language.

🌤️ Want to skip the setup? Run massive multi-omics pipelines instantly on Omics-OS Cloud!

---

🧑‍🔬 Human Quickstart

1. Install Lobster AI (macOS/Linux):

curl -fsSL https://install.lobsterbio.com | bash

(Windows users: irm https://install.lobsterbio.com/windows | iex)

2. Configure your LLM (Anthropic, Gemini, local Ollama, etc.):

lobster init

3. Run your first analysis:

lobster query "Download GSE109564 and cluster the cells"
# Or run interactively: lobster chat

🤖 For AI Coding Agents

Teach your coding agent (Claude Code, Cursor, Gemini) to use and extend Lobster AI instantly:

curl -fsSL https://skills.lobsterbio.com | bash

Installs the lobster-use and lobster-dev skills so your AI knows our entire 10-package architecture.

🎬 Watch it Work

🧬 Single-Cell Transcriptomics

Task: "Download GSE109564, run QC, and cluster cells."

$ lobster query "Download GSE109564, run QC, and cluster cells"
[Supervisor] Routing to Transcriptomics Expert...
[Transcriptomics] Delegating download of 'GSE109564' to Data Expert...
[Data Expert] Download complete. Modality 'GSE109564' loaded.
[Transcriptomics] Running QC (calculating MT fraction, filtering cells/genes)...
[Transcriptomics] Normalizing and finding highly variable genes...
[Transcriptomics] Running PCA, Neighborhood graph, and UMAP...
[Transcriptomics] Clustering via Leiden (resolution=1.0)...
✅ Analysis complete. UMAP plot generated and saved to workspace.

🔬 Mass Spec Proteomics

Task: "Import MaxQuant data, perform batch correction, and select biomarker panels."

$ lobster query "Import MaxQuant data, perform batch correction, and select biomarker panels"
[Supervisor] Routing to Proteomics Expert...
[Proteomics] Parsing proteinGroups.txt and experimental design...
[Proteomics] Filtering contaminants and reverse hits...
[Proteomics] Performing median normalization and ComBat batch correction...
[Proteomics] Delegating to Biomarker Discovery Expert...
[Biomarker Expert] Running LASSO stability selection with nested cross-validation...
✅ Complete. 12 robust biomarker candidates identified (Stability score > 0.8).

📚 Automated Literature Discovery

Task: "Search PubMed for CRISPR studies in 2024 and download the top 3 datasets."

$ lobster query "Search PubMed for CRISPR studies in 2024 and download the top 3 datasets."
[Supervisor] Routing to Research Agent...
[Research] Searching PubMed: "(CRISPR[Title/Abstract]) AND 2024[Date - Publication]"
[Research] Extracting GEO accessions from top 10 relevant papers...
[Research] Found GSE251842, GSE252910, GSE260124.
[Research] Delegating batch download to Data Expert...
✅ Datasets downloaded and harmonized into AnnData objects.

🧠 The Architecture

Lobster isn't just a chatbot; it's a modular ecosystem of 22 specialist agents across 10 packages.

• Your machine, your data: Patient data never leaves your hardware.
• Tool calls, not token dreams: Agents execute real, validated Python packages (Scanpy, PyDESeq2).
• 100% Reproducible: W3C-PROV tracking and automatic Jupyter notebook exports.

🛠️ Build Your Own Agent

The lobster-dev skill gives your coding assistant (Claude Code, Gemini CLI, Cursor) deep knowledge of how Lobster agents are structured. Describe the biological domain you need — it scaffolds the package, wires the tools, writes the tests, and registers the agent.

1. The Request

2. The Result

❓ Deep Dives & FAQ

What omics domains are supported?

Transcriptomics

• Single-cell RNA-seq: QC, doublet detection (Scrublet), batch integration (Harmony/scVI), clustering, cell type annotation, trajectory inference (DPT/PAGA)
• Bulk RNA-seq: Salmon/kallisto/featureCounts import, sample QC, batch detection, normalization (DESeq2/VST/CPM), DE with PyDESeq2, GSEA, publication-ready export

Genomics

• GWAS: VCF/PLINK import, LD pruning, kinship, association testing, result clumping
• Clinical: variant annotation (VEP), gnomAD frequencies, ClinVar pathogenicity, variant prioritization

Proteomics

• Mass spec: MaxQuant/DIA-NN/Spectronaut import, PTM analysis, peptide-to-protein rollup, batch correction
• Affinity: Olink NPX/SomaScan ADAT/Luminex MFI import, LOD quality, bridge normalization
• Downstream: GO/Reactome/KEGG enrichment, kinase enrichment (KSEA), STRING PPI, biomarker panel selection

Metabolomics

• LC-MS, GC-MS, NMR with auto-detection
• QC (RSD, TIC), filtering, imputation, normalization (PQN/TIC/IS)
• PCA, PLS-DA, OPLS-DA, m/z annotation (HMDB/KEGG), lipid class analysis

Machine Learning

• Feature selection (stability selection, LASSO, variance filter)
• Survival analysis (Cox models, Kaplan-Meier, risk stratification)
• Cross-validation, SHAP interpretability, multi-omics integration (MOFA)

Research & Metadata

• Literature discovery (PubMed, PMC, GEO, PRIDE, MetaboLights)
• Dataset download orchestration, metadata harmonization, sample filtering

Which LLMs can I use?

Lobster supports 5 LLM providers. Configure via lobster init or environment variables.

Provider	Type	Setup	Use Case
Ollama	Local	ollama pull gpt-oss:20b	Privacy, zero cost, offline
Anthropic	Cloud	API key	Fastest, best quality
AWS Bedrock	Cloud	AWS credentials	Enterprise, compliance
Google Gemini	Cloud	Google API key	Multimodal, long context
Azure AI	Cloud	Endpoint + credential	Enterprise Azure

Pipeline export and slash commands

lobster chat
> /pipeline export         # Export reproducible Jupyter notebook
> /pipeline list           # List exported pipelines
> /pipeline run analysis.ipynb geo_gse109564
> /data                    # Show loaded datasets
> /status                  # Session info
> /help                    # All commands

Advanced installation (Windows, pip)

Windows (PowerShell):

irm https://install.lobsterbio.com/windows | iex

uv (recommended manual install):

uv tool install 'lobster-ai[full,anthropic]'
lobster init

pip:

pip install 'lobster-ai[full]'
lobster init

Upgrade:

uv tool upgrade lobster-ai    # uv
pip install -U lobster-ai      # pip

How do I build my own agent?

Create custom agents for any domain. Agents plug in via Python entry points — discovered automatically, no core changes needed.

Install the lobster-dev skill to teach your coding agent the full architecture:

curl -fsSL https://skills.lobsterbio.com | bash

Then ask your coding agent: "Create a Lobster agent for [your domain]" — it knows the package structure, AGENT_CONFIG pattern, factory function, tool design, testing, and the 28-step checklist.

Built to accelerate multi-omics research.

Omics-OS  ·  Lobster AI  ·  Docs