P PrimerCode / omicverse
EN·中
Launch App→

The agent operating system
for biological discovery.

Specialist agents read your data, plan analyses, and write reproducible notebooks — powered by 1,000+ omicverse functions and your choice of frontier model.

Launch Web App→ Get API partnership→
Compatible with
OpenAI Anthropic Google Gemini DeepSeek Qwen Local LLMs
Agents

Specialist agents that know omics.

Ten purpose-built agents share the same kernel and 1,000+ omicverse functions. Switch the underlying model per turn — chat history and variables stay synced.

◆
Leader
orchestrator
Routes user goals to the right specialist; runs simple one-shots directly.
Tools
call_agent python_interpreter file_manager
⊟
Loader
data ingest
Loads .h5ad / .csv / .mtx / 10x outputs into a shared kernel as AnnData.
Tools
file_manager python_interpreter omicverse_lookup
▣
QC
quality control
Mitochondrial %, doublet detection, gene/cell filters with sensible defaults.
Tools
python_interpreter omicverse_lookup
≣
Preprocess
normalize · HVG
Total-count normalization, log1p, HVG selection, scaling to unit variance.
Tools
python_interpreter omicverse_lookup
⊕
Integrate
batch correction
Harmony / scVI / BBKNN batch integration with auto-selected method per dataset.
Tools
python_interpreter omicverse_lookup
∴
Clustering
leiden · UMAP
Neighbors graph, Leiden / Louvain partitioning, UMAP / t-SNE embeddings.
Tools
python_interpreter omicverse_lookup
◉
Annotation
cell type labels
CellTypist + scANVI + LLM-based annotation; cross-checks methods, flags ambiguity.
Tools
python_interpreter omicverse_lookup file_manager
↗
Trajectory
pseudotime
Diffusion map, Slingshot, Palantir, scTour pseudotime; PAGA overlay on embedding.
Tools
python_interpreter omicverse_lookup
⊞
Spatial
visium · merfish
STAGATE clustering, Tangram deconvolution, multi-slide alignment.
Tools
python_interpreter omicverse_lookup
✦
Expert
consultation
Free-form deep dives — interprets results, suggests follow-ups, drafts methods sections.
Tools
python_interpreter omicverse_lookup
Product

Talk to your data. Agents do the rest.

A web workbench where chat, notebook, and a shared Python kernel live side-by-side. Drop an .h5ad, ask a question — the right specialist picks up the work.

OmicOS Web app showing chat, notebook and active agents

Drag-and-drop ingest

.h5ad, 10x mtx, AnnData URLs. Variables auto-detected and surfaced to every agent.

Multi-agent collaboration

Specialists hand off to each other; the Collaboration pane shows the flow in real time.

Notebook + chat synced

Every tool call lands in a Jupyter notebook you can run, edit, and export.

Engine

The omicverse Python library.

A unified framework for bulk RNA-seq, single-cell, and spatial transcriptomics — 1,000+ functions across 9 modules, with a discoverable function registry that agents can search at runtime.

$ pip install omicverse copy
Read the docs↗ View on GitHub↗
Ecosystem

21 packages.
One coherent API.

A flagship library, an agent SDK, drop-in Python ports of canonical R / Bioconductor tools, and Rust accelerators where it counts.

Core

omicverse
★ 1,000+
Unified Python library for bulk, single-cell, and spatial transcriptomics. The engine layer.
GitHub Docs PyPI
omicclaw
★ 27
Agent-skill plugin for Claude Code — gives any LLM IDE direct access to omicverse.
GitHub
omicverse-skills
★ 2
Agent-toolset package — the runtime registry every OmicOS agent reads.
GitHub PyPI
anndata-oom
★ 2
Out-of-memory AnnData powered by Rust — drop-in replacement that keeps X on disk.
GitHub PyPI

Pure-Python ports of canonical R / Bioconductor tools

py-monocle2→Monocle 2 — trajectory inferenceQiu et al., Nature Methods 2017
py-miloR→Milo — DA testing on kNN graphsDann et al., Nat. Biotech. 2022
py-mclustR→CRAN mclust — Gaussian mixture clusteringScrucca et al., 2016
py-DoubletFinder→DoubletFinder — doublet detectionMcGinnis et al., Cell Systems 2019
py-scDblFinder→scDblFinder — Bioc doublet detectionGermain et al., F1000Research 2021
py-dada2→DADA2 — ASV inference from ampliconsCallahan et al., Nature Methods 2016
py-cca→Seurat RunCCA — single-cell integrationStuart et al., Cell 2019
py-CopyKAT→CopyKAT — aneuploidy & CNA from scRNAGao et al., Nat. Biotech. 2021
py-scmetabolism→scMetabolism — single-cell metabolismWu et al., Nat. Protocols 2022
py-inferCNV→inferCNV — single-cell CNV inferenceBroad Institute

Rust performance ports

rust-scHiCluster
~10× faster
Rust port of the hot loops in scHiCluster (Zhou et al., PNAS 2019). Bit-identical numerics at published defaults.
GitHub
rust-bandnorm
~24× faster
Rust port of BandNorm scGAD + bandnorm normalisation for single-cell Hi-C.
GitHub

Learn

single_cell_tutorial
★ 286
单细胞分析中文 pipeline — tutorial collection in Mandarin.
GitHub
omicverse-tutorials
★ 3
Main tutorial collection — bulk, single-cell, spatial recipes.
GitHub
omicverse-to-developer
★ 0
How to port an R/Bioconductor package to Python and integrate it back.
GitHub
Extend

Build your own agents.

Your model + our agents = a vertical product. Both the agent SDK and the toolset registry are open and pip-installable.

omicclaw

Drop-in skill for any agent IDE — Claude Code, Cursor, Continue. Inherits the full omicverse function registry.

View on GitHub →

omicverse-skills

The toolset registry every OmicOS agent reads at runtime. Add a new tool, get it across all 10 specialists.

View on GitHub →
1,000+ functions · 10 specialist agents · 21 packages · Nature Communications 2024
Latest

Writing & releases

View all →
2026-04 OmicOS launches: chat-driven multi-omics analysis Product
2026-03 Pure-Python port of Monocle 2: a drop-in replacement Engineering
2026-02 v1.7.9: 70+ tool integrations and a unified function registry Release
2026-01 24× faster Hi-C with rust-scHiCluster Engineering
2025-12 Nature Communications 2024 — the omicverse paper Research
For AI Labs

API partnership.

If you're building frontier models and want a vertical foothold in biology, talk to us. We're model-agnostic and bring 1,000+ tool calls, 10 specialist agents, and a real research user base.

PrimerCode
Founded 2026 · 源境解码

We're building open infrastructure for next-generation biological discovery. omicverse is the engine; OmicOS is the agent host. Together, they're the substrate for a generation of AI-native biology products — including ours, and yours.