Platform Ecosystem

Open-source tooling for
causal builders

Causal-Copilot sits beneath the Abel Platform as the open-source toolkit for causal discovery workflows.

The most comprehensive open-source causal discovery library. Created by Biwei Huang (Abel co-founder, UC San Diego Professor, creator of causal-learn). The general backend of Abel's Social Physical Engine.

Abel runs GPU-accelerated PCMCI — a core algorithm in this library — daily across 200K+ variables to power its live causal world model.

39
Discovery Algorithms
6
Inference Methods
11
Selection Dimensions
200+
Benchmark Scenarios

Algorithm Library

The most comprehensive causal discovery toolkit

Organized by methodology. Abel's algorithm selector automatically chooses the best method for your data.

Constraint-Based

Test conditional independence to discover causal structure.

PCFCICPCFCI+RFCICD-NODPCMCIPCMCI+LPCMCI

Score-Based

Search for the causal graph that best fits the data.

GESFGESXGESBOSSGRASPExactSearch

Functional

Exploit functional asymmetries to determine causal direction.

LiNGAMDirectLiNGAMVarLiNGAMRESITANMPNL

Hybrid & Other

Combine constraint and score methods, or use deep learning.

NOTEARSGOLEMGraNDAGDAG-GNNNTS-NOTEARSDYNOTEARSCDNODACD

Causal Inference

6 methods for estimating causal effects

Once you discover the graph, use these methods to estimate how much X causes Y.

DMLDouble Machine Learning

Semiparametric estimation of causal effects with machine learning nuisance models.

DRLDoubly Robust Learning

Robust to misspecification of either outcome or treatment model.

IVInstrumental Variables

Handle unmeasured confounders using instrumental variable estimation.

MetaLearnersS/T/X-Learners

Estimate heterogeneous treatment effects across subpopulations.

MatchingPropensity Score Matching

Match treated/control units on propensity scores for effect estimation.

WeightingIPW / AIPW

Inverse probability weighting for unbiased causal effect estimation.

Automatic Selection

11-dimensional algorithm selector

Don't know which algorithm to use? Neither does the average data scientist. Abel's selector benchmarks 200+ scenarios and picks the optimal method for your data characteristics.

Sample size
Dimensionality
Linearity
Stationarity
Noise type
Missing data
Latent confounders
Time series
Mixed data types
Graph density
Domain constraints

Used by researchers worldwide

Created by the person who built the tools the field uses. Causal-Copilot powers research at top universities and is the foundation of Abel's $2M+/year GPU infrastructure.

UC San DiegoStanfordMITCMUBerkeleyPrinceton

Start making decisions with
live causal intelligence.

Interested in shaping the future of
causal intelligence? We're hiring.

Open Roles

Join Abelian Groups to stay on top of new
releases, features, and updates.