⭐ What's new in 0.14.0
Interventional data: rows from experiments, where a variable was changed from outside. Where
observation leaves the direction of an edge open, an experiment on either end can settle it. IGSP
learns graphs from experiments with known targets, GnIES and UT-IGSP from experiments whose targets
they are not told, and ICP finds the causes of one variable, all as accurately as their authors'
code. CAM-UV and DYNOTEARS are supported too.
Read the post ·
See the release card
Andrey is a causal discovery package for Python.
- Fast: The fastest causal discovery package.
- Accurate: more than 80% fewer errors than other popular packages.
- One API: 40 supported methods, each one function call.
- Any data: continuous, categorical, mixed, time series, nonlinear, hidden confounders.
- Agent-ready: a command line that answers in JSON, and a skill for coding agents.
uv add andrey-core
pip install andrey-core
You install andrey-core and then import andrey.
- Optional extras:
andrey-core[torch] for GPUs, andrey-core[viz] for charts and graph
layouts
- Development version:
uv add "andrey-core @ git+https://github.com/Abel-ai-lab/andrey@main"
- Installation guide: the
extras, a PyTorch build for your GPU, and building Andrey from source
What's inside
| Family | Methods | Returns |
|---|
| Constraint-based | pc, fci, gfci, cdnod, icp, iamb, inter_iamb, mmpc | CPDAG, PAG, DAG, UNDIRECTED |
| Score-based | ges, gies, gnies, xges, hc, tabu, exact_search, , *, *, , |
* Experimental:
no published benchmark yet, and the API may change.
Quick start
Learn the protein-signaling network of Sachs et al. (2005) from 853 cells, and score it against the
known network:
import andrey
from andrey.data import load_dataset
from andrey.metrics import score
sachs = load_dataset("sachs") # 853 cells, 11 proteins, and the known network
out = andrey.pc(sachs.data) # learn a graph
print(out)
scores = score(out.structure, sachs.graph)
print({k: round(scores[k], 2) for k in ("shd", "skeleton_precision", "skeleton_recall")})
pc cpdag | 11 nodes | 8 edges (6 undirected, 2 directed)
raf -- mek
plc -- pip3
pip2 -- pip3
erk -- akt
erk -- pka
akt -- pka
p38 -> pkc
jnk -> pkc
indep_test: fisherz
tie_break: rank
{'shd': 11, 'skeleton_precision': 1.0, 'skeleton_recall': 0.47}
PC finds 8 edges, all among the known network's 17. The structural Hamming distance (SHD) counts
the edges to add, remove, or reorient to reach the known network: 11 here (9 edges are missing, and
2 are oriented that the known network leaves undirected), and 17 for a graph with no edges.
For your own data, pass a pandas DataFrame or a NumPy array: one row per sample, one column per
variable. From a shell or an agent:
andrey run pc --data your_data.csv # JSON when piped, a summary in a terminal
To give your coding agent Andrey's skill, paste this prompt into it:
Install Andrey's agent skill: make sure the andrey-core package is installed, run
`andrey --skill`, and save its output as andrey/SKILL.md in your skills folder.
Learn more
Andrey is in alpha: the API may change between releases, so pin the version you use. Linux is fully
tested; macOS and Windows install and run on the CPU, but the full test suite and the GPU speedups
(CUDA / MPS) are not tested there yet.
Citing, license, contributing
If you use Andrey in your work, please cite it, or use the
citation file:
@software{andrey,
author = {{Abel AI Lab}},
title = {Andrey: Accelerated, accurate, agentic causal discovery},
year = {2026},
version = {0.14.0},
url = {https://andrey.abel.ai},
}
Andrey is released under the
Apache License 2.0. Bug reports and pull
requests are welcome in the issue tracker; the
contributing guide explains how
to set up a development environment.
Andrey is inspired by
causal-learn, pgmpy,
Tetrad, and many other causal discovery packages.
