Overview
is a closed-loop framework for evidence-grounded neuroimaging autoresearch. It comprises a , a and , an agent-based execution platform. The graph combines curated resources with source-linked literature evidence, preserving provenance, publication dates and evidence polarity. The generator constructs testable cross-domain hypotheses compatible with available data and analysis methods. NeuroRuntime tests selected hypotheses reproducibly on raw neuroimaging data; supported, contradicted and inconclusive outcomes guide subsequent research, separately from execution failures. NeuroRuntime retains the project's strengths in , and . It ships with independent GUI and CLI interfaces for day-to-day use, and can also be installed as a reusable skill library inside agent projects such as OpenClaw, Hermes, and Claude Code. This repository hosts (formerly ) and related public resources.
README
📖 Overview
NeuroDiscovery is a closed-loop framework for evidence-grounded neuroimaging autoresearch. It comprises a neuroscience knowledge graph, a hypothesis generator and NeuroRuntime, an agent-based execution platform.
The graph combines curated resources with source-linked literature evidence, preserving provenance, publication dates and evidence polarity. The generator constructs testable cross-domain hypotheses compatible with available data and analysis methods. NeuroRuntime tests selected hypotheses reproducibly on raw neuroimaging data; supported, contradicted and inconclusive outcomes guide subsequent research, separately from execution failures.
NeuroRuntime retains the project’s strengths in neuroimaging dataset and model adaptation, data processing and model configuration/execution. It ships with independent GUI and CLI interfaces for day-to-day use, and can also be installed as a reusable skill library inside agent projects such as OpenClaw, Hermes, and Claude Code.
NeuroDiscovery and research materials
This repository hosts NeuroDiscovery (formerly NeuroClaw) and related public resources. Existing directory and interface names, including neurooracle/, neurobench/, neuroclaw_environment.json, the neuroclaw host-agent skill and neuroclaw_academic_* MCP tools, are retained for compatibility.
Public materials have separate versions:
| Resource | Availability |
|---|---|
| Neuroimaging execution tasks | 500 task definitions, T01–T500, with a complete seven-category task registry. See the benchmark README. |
| Knowledge-graph explorer | The NeuroOracle demo provides interactive graph exploration. Its loaded snapshot can differ from the research graph. See the graph version notes. |
| Benchmark outputs | Selected historical evaluation outputs are available. Their task coverage is recorded per run. |
| NeuroDiscovery manuscript release | The frozen graph, paper-specific evaluation outputs and figure source data are being prepared for a versioned research release. A release identifier and artifact manifest will be linked here when available. |
The linked NeuroClaw technical report describes an earlier project version. Its experiments and the public demo should be interpreted using their own version information, separately from the NeuroDiscovery manuscript.
🚀 Updates
- [2026.09.15]: NeuroDiscovery v1.0.0 brings a refreshed desktop interface, expanded model-provider support including Ollama Cloud, improved AutoResearch continuity, and three configurable hypothesis novelty modes (
strict,novelty_first,weighted). - [2026.06.20]: NeuroClaw now provides Windows and macOS desktop clients, while Linux remains supported through the repository and command-line/web workflows.
- [2026.05.23]: NeuroBench now covers both data processing and model training/evaluation.
- [2026.05.20]: 7 atoms × 15 canonical tasks + 4 mediation chains in
neurooracle.atoms. - [2026.05.15]: NeuroOracle launched: knowledge-graph explorer plus hypothesis engine with live demo at https://huggingface.co/spaces/zxcvb20001/NeuroOracle.
- [2026.05.06]: Added 19 dataset and modality skills with companion scripts; all 86 skills enforce unified metadata (
layer,skill_type,dependencies); skill_loader DAG validation ensures dependency graph correctness. - [2026.04.28]: Our technical report is now available on arXiv: https://arxiv.org/abs/2604.24696
- [2026.04.22]: v1.0 released. Stable release with improvements and full documentation.
- [2026.04.17]: Our project homepage is now live. Welcome to visit: https://cuhk-aim-group.github.io/NeuroDiscovery/
- [2026.04.08]: NeuroBench released for multi-agent neuroimaging workflow evaluation.
- [2026.04.02]: v0.1 released with complete NeuroClaw framework and core functionality.
✨ Key Features
🔄 Data-Aware Orchestration
- Dataset-Context Planning: Organize capabilities around dataset structure, metadata, and workflow stage instead of simply “which tool to call”
- Automatic Skill Recommendation: Users specify the target dataset, and NeuroRuntime recommends relevant skills and executable workflows
- Preprocessing Constraint Awareness: Dataset-specific modality availability and preprocessing requirements are considered during orchestration
Supported Dataset Overview
Access is not equivalent to anonymous download. See the verified access matrix for registration, DUA, review, fee, and current availability details.
🎯 Executability and Reproducibility
- Automatic Dependency Management: No manual installation needed; the system detects and resolves dependencies
- True Model Execution: Beyond sharing docs, it guides and executes model reproduction
- Environment Isolation: Virtual environments and containerization avoid system pollution
- Verifiable Processes: Complete logging and result tracking
- Shadow Checkpoints: Git-based filesystem snapshots for rollback and diff comparison without polluting the project repository
- Subagent Orchestration: Spawns specialized subagents (biostatistician, clinical neuroscientist, methodology expert) for multi-perspective task execution
- Reflective Learning: Automatic reflection on tool failures and task completion, with persistent memory for cross-session learning
🧠 End-to-End Research Coverage
- Literature Review: arXiv search, PubMed retrieval, academic resource integration
- Experiment Design: Evidence-grounded hypothesis generation, scientific literature analysis and methodology evaluation
- Data Processing: Multi-format conversion (DICOM ↔ NIfTI), automated preprocessing pipelines
- Model Execution: Run published research models, deep learning framework integration
- Result Visualization: Scientific data visualization, statistical chart generation
- Paper Writing: Auto-generated drafts, format standardization
🤝 Flexible Integration
- NeuroDiscovery provides standalone GUI and CLI workflows through NeuroRuntime, so researchers can use it directly without depending on another host project.
skills/,materials/,USER.md, andSOUL.mdcan also be installed as a reusable skill library in existing agent systems such as OpenClaw, Hermes, and Claude Code.- The bundled
core/engine provides an integrated agent loop, skill loader, and tool runtime for standalone deployments. - Non-neuroscience connectors (WhatsApp, Telegram, Slack, calendar, e-commerce, SaaS auth)
are disabled by default via
core/config/features.jsonand can be re-enabled if needed.
🚀 Quick Start
Option 1. Desktop Client (recommended)
Download the latest Windows or macOS client from the GitHub Releases page. Existing release assets retain their original NeuroClaw filenames.
- Windows: use
NeuroClaw Setup 0.2.1.exefor normal installation. The portable.exeis also available, but may take longer to start because it extracts the app first. - macOS: use the
.dmgor.zipbuild from the release assets. - Open Settings to configure the model endpoint, runtime mode, Python path, FSL path, proxy, language, and text size.
- Open NeuroOracle from the sidebar. If the graph file is missing, the client can download it from Hugging Face.
Linux remains supported through the source repository, command-line workflow, and web interface.
Option 2. Run from Source
Requirements: Python >= 3.10 and Git. Conda/Mamba, CUDA/GPU tools, FSL, FreeSurfer, and dcm2niix are optional depending on the workflows you want to run.
git clone https://github.com/CUHK-AIM-Group/NeuroDiscovery.git
cd NeuroDiscovery
python installer/setup.py
python core/agent/main.py --web
Then open http://localhost:7080 in your browser.
Useful checks:
python installer/setup.py --check
python core/agent/main.py --web --port 8080 --host 0.0.0.0
Settings are saved to neuroclaw_environment.json. API keys can be passed at runtime with --api-key or provided through the configured provider environment variable.
Option 3. Install as a Host-Agent Skill
Use this path if you want Codex, Claude Code, Cursor, or another coding agent to use NeuroDiscovery’s neuroimaging skill library.
git clone https://github.com/CUHK-AIM-Group/NeuroDiscovery.git
cd NeuroDiscovery
python installer/install_agent_integration.py --target codex
Common targets:
| Host agent | Install command |
|---|---|
| Codex | python installer/install_agent_integration.py --target codex |
| Claude Code | python installer/install_agent_integration.py --target claude-code |
| Cursor | python installer/install_agent_integration.py --target cursor --scope project |
| Multiple agents | python installer/install_agent_integration.py --target all |
The installed host-agent skill retains its compatibility name, neuroclaw. After installation, ask the host agent to use NeuroClaw or enter NeuroClaw mode for neuroimaging, NeuroOracle, NeuroBench, and autoresearch tasks.
Benchmark output files under
materials/benchmark_results/are historical run artifacts. See their coverage and scoring notes before comparing them with a newer task registry.
Benchmark Evaluation
The 500 neuroimaging execution tasks live under neurobench/. Each task directory contains a task.md instruction file, and task_atlas.json assigns every task to one of seven categories. NeuroBench remains the name used by the existing benchmark interface.
NeuroBench currently accepts these benchmark configurations:
with-skills: the agent can use the skills loaded fromskills/no-skills: the baseline run without skillswith-skills+no-skillspaired comparison: enable--benchmark-compare-skillsto run both variants for the same task set
Benchmark scoring is handled separately with --score-benchmark: it reads reports in output/, applies a GPT-5.4 weighted rubric, and generates numeric scores for planning completeness, tool/skill reasonableness, and command/code correctness. For fairness, each task case is scored in one batch across all comparable models to reduce scoring-standard drift. Skill-call counts are recorded separately and used for efficiency analysis.
To score existing benchmark reports:
python core/agent/main.py --score-benchmark
To speed up scoring on larger runs:
python core/agent/main.py --score-benchmark --score-workers 8
Web benchmark mode
python core/agent/main.py --web --benchmark
CLI benchmark batch runner
python core/agent/main.py --benchmark
To run the paired skill comparison in CLI mode:
python core/agent/main.py --benchmark --benchmark-compare-skills
In CLI benchmark mode, NeuroRuntime will ask for:
- the benchmark directory path
- the benchmark model name
Then it will:
- read all
task.mdfiles recursively from that directory - sort tasks alphabetically by task folder name
- run tasks one by one without asking for intermediate confirmation
- print progress in the terminal only
- save reports under
output//, with one markdown report per case and run
The benchmark reports include the solution thinking, skills used, skill-call counts, and the commands or code that were used or suggested.
📁 Project Structure
NeuroDiscovery/
├── README.md / README_zh.md # Project documentation
├── USER.md / SOUL.md # User preferences and agent behavior guidelines
│
├── core/ # NeuroRuntime execution platform
│ ├── agent/ # CLI/Web agent entry points
│ ├── web/ # FastAPI Web UI
│ ├── skill_loader/ # Reads skills/*/SKILL.md
│ └── config/ # Feature toggles and runtime settings
│
├── installer/ # Setup wizard and host-agent integration installer
│ ├── setup.py
│ ├── config_wizard.py
│ └── install_agent_integration.py
│
├── skills/ # Skill library
│ ├── base skills # Environment, search, BIDS, Git, conversion
│ ├── interface skills # Research idea, method design, experiments, writing
│ └── subagent skills # Tool, model, dataset, and modality workflows
│
├── models/ # Brain model adapters and training/evaluation scripts
├── neurooracle/ # Knowledge graph and autoresearch pipeline
│
├── neurobench/ # 500 neuroimaging execution tasks (T01-T500)
│
├── docs/ # Project website pages
├── materials/ # Research materials and benchmark outputs
│
└── LICENSE # License
🛠️ Skill Quick Reference
Tip: Click the ℹ️ icon on any skill card in the Web UI to view expanded documentation, usage examples, and recent execution logs.
Base Layer
| Skill | Function | Status |
|---|---|---|
dcm2nii |
DICOM → NIfTI conversion with metadata support | ✅ |
nii2dcm |
NIfTI → DICOM conversion for clinical interoperability | ✅ |
git-essentials |
Core Git commands for collaboration | ✅ |
git-workflows |
Advanced Git workflows (rebase/worktree/bisect) | ✅ |
multi-search-engine |
Multi-engine web search without API keys | ✅ |
conda-env-manager |
Conda environment lifecycle management | ✅ |
docker-env-manager |
Docker environment management | ✅ |
dependency-planner |
Dependency planning and safe installation workflow | ✅ |
claw-shell |
Safe shell execution gateway via dedicated session | ✅ |
overleaf-skill |
Overleaf sync and collaborative manuscript operations | ✅ |
academic-research-hub |
Multi-source academic search and paper retrieval | ✅ |
bids-organizer |
Base skill for organizing raw data into BIDS structure | ✅ |
beautiful-log |
Export clean User/NeuroDiscovery dialogue into beautiful HTML logs | ✅ |
knowledge-graph-builder |
Build domain knowledge graphs from literature and databases | ✅ |
skill-updater |
Skill updater and management utilities | ✅ |
Interface Layer (Task Orchestration)
| Skill | Function | Status |
|---|---|---|
research-idea |
Brainstorms and generates research ideas from literature | ✅ |
method-design |
Formalizes network architecture and derives theoretical components | ✅ |
experiment-controller |
Finds and executes reproducible research experiments | ✅ |
paper-writing |
Generates hierarchical manuscript drafts from IDEA/METHOD/EXPERIMENT | ✅ |
Subagent Layer
Subagent skills in NeuroRuntime include four categories: tool, model, dataset, and modality.
Tool
| Skill | Function | Status |
|---|---|---|
brain-visualization |
Publication-ready figures and 3D assets (connectomes, atlas summaries, FreeSurfer PLY) | ✅ |
harmonization-tool |
Cross-site / cross-scanner feature harmonization (ComBat, ComBat-GAM, CovBat, site-as-covariate) with site-stratified and leave-site-out splitters; required for honest mega-analysis across multi-site cohorts | ✅ |
harness-core |
Core harness SDK: verification, checkpointing, drift detection, audit logging | ✅ |
mne-eeg-tool |
Base-layer MNE-Python implementation for EEG | ✅ |
fsl-tool |
FSL-based sMRI/fMRI/DWI processing utilities | ✅ |
fmriprep-tool |
fMRIPrep pipeline wrapper and execution | ✅ |
qsiprep-tool |
qsiPrep pipeline wrapper for diffusion MRI | ✅ |
hcppipeline-tool |
HCP-style processing pipeline utilities | ✅ |
dipy-tool |
Diffusion MRI processing via DIPY | ✅ |
nibabel-skill |
Low-level neuroimaging I/O and geometry handling (NIfTI, affine, FreeSurfer I/O) | ✅ |
nilearn-tool |
Fast neuroimaging feature extraction and decoding prep | ✅ |
conn-tool |
Functional connectivity computation and analysis | ✅ |
freesurfer-tool |
FreeSurfer-based MRI processing and segmentation | ✅ |
Model
| Skill | Function | Status |
|---|---|---|
run_models |
Model registry and model execution orchestration | ✅ |
wmh-segmentation |
White matter hyperintensity segmentation (MARS-WMH nnU-Net) | ✅ |
brain_gnn |
BrainGNN: graph neural network for fMRI classification | ✅ |
bnt |
BrainNetworkTransformer: dense FC Transformer with DEC pooling for phenotype prediction | ✅ |
brainnetcnn |
BrainNetCNN: E2E/E2N/N2G convolutions over dense connectivity matrices | ✅ |
combraintf |
Com-BrainTF: community-aware two-level Transformer over dense FC matrices | ✅ |
ibgnn |
IBGNN: interpretable PyG-based GNN with MLP message function and edge-mask explainer | ✅ |
lggnn |
LG-GNN: PyG-based GNN with Self-Attention Brain Pooling and mutual-information regularization | ✅ |
fm_app |
FM-APP: multi-stage phenotype prediction with fMRI+sMRI | ✅ |
neurostorm |
NeuroStorm: neuroimaging foundation model | ✅ |
glm |
Classical first-level and second-level GLM for task-fMRI activation and group inference | ✅ |
ica |
Resting-state network decomposition via independent component analysis | ✅ |
dictlearning |
Sparse resting-state network decomposition via dictionary learning | ✅ |
spacenet |
Voxel-wise neuroimaging disease classification with sparse coefficient maps | ✅ |
kmeans |
Brain parcellation via K-means clustering | ✅ |
hierarchical |
Multi-scale brain parcellation via hierarchical clustering | ✅ |
filtering |
Temporal filtering for neuroimaging signal denoising | ✅ |
detrending |
Temporal drift removal for neuroimaging signal denoising | ✅ |
statistical-ml |
Unified tabular OLS/GLM, SVM/SVR, Ridge, Elastic Net, XGBoost, and mixed-effects models | ✅ |
subject-subtyping |
Subject-level subtyping with clustering and latent embeddings | ✅ |
survival-models |
Censor-aware Cox, RSF, DeepSurv, and XGBoost survival models | ✅ |
causal-treatment-models |
Cross-fitted treatment-effect and individualized policy models | ✅ |
temporal-models |
LSTM, GRU, TCN, and temporal Transformer sequence models | ✅ |
imaging-genetics-models |
Association, LMM, PRS, PLS, and CCA imaging-genetics models | ✅ |
cnn3d |
Compact residual 3D CNN for voxel-level prediction | ✅ |
cpm |
Connectome Predictive Modeling with fold-local edge selection | ✅ |
kg-link-prediction |
ComplEx, R-GCN, GraphSAGE, and GAT knowledge-graph link prediction | ✅ |
Workflow
| Skill | Function | Status |
|---|---|---|
neuroimaging-decoding |
Coordinates ROI MVPA, ROI GLM, and voxel-wise SearchLight analysis | ✅ |
connectome-discovery |
Converts connectome-model outputs into significant maps and ranked targets | ✅ |
brain-age-modeling |
Cross-validated brain-age prediction with fold-local bias correction | ✅ |
Dataset
| Skill | Function | Status |
|---|---|---|
abide-skill |
ABIDE dataset download, BIDS staging, and sMRI/rs-fMRI processing | ✅ |
aibl-skill |
AIBL dataset access, BIDS staging, and sMRI/PET processing | ✅ |
abcd-skill |
ABCD Study controlled NBDC access, BIDS staging, and multimodal processing | ✅ |
adhd200-skill |
ADHD-200 dataset download, BIDS staging, and sMRI/rs-fMRI processing | ✅ |
adni-skill |
ADNI and ADNI-DOD controlled access, BIDS staging, and processing workflow | ✅ |
aomic-skill |
AOMIC dataset validation, BIDS staging, and sMRI/rs-fMRI/task-fMRI processing | ✅ |
bold5000-skill |
BOLD5000 dataset BIDS validation and visual task-fMRI processing | ✅ |
camcan-skill |
Cam-CAN dataset BIDS validation, multimodal sMRI/rs-fMRI/task-fMRI/dMRI processing | ✅ |
cobre-skill |
COBRE dataset BIDS staging and schizophrenia-control fMRI processing | ✅ |
dmt-har-med-skill |
DMT-HAR-MED dataset BIDS validation and psychedelic rs-fMRI processing | ✅ |
hbn-skill |
HBN dataset download, BIDS staging, and multimodal sMRI/fMRI/dMRI/EEG processing | ✅ |
hcpa-skill |
HCP Aging/AABC access, BIDS staging, and multimodal sMRI/fMRI/dMRI/ASL processing | ✅ |
hcpd-skill |
HCP Development dataset download, BIDS staging, and multimodal sMRI/fMRI/dMRI processing | ✅ |
hcpep-skill |
HCP Early Psychosis dataset download, BIDS staging, and multimodal sMRI/fMRI/dMRI processing | ✅ |
hcpya-skill |
HCP Young Adult 2025/S1200 access, BIDS staging, and multimodal sMRI/fMRI/dMRI processing | ✅ |
ixi-skill |
IXI dataset BIDS validation and multimodal sMRI/MRA/dMRI processing | ✅ |
mnd-skill |
MND dataset BIDS validation, rs-fMRI/task-fMRI processing, and phenotype extraction | ✅ |
mschallenge-skill |
MS Lesion Challenge BIDS validation, lesion analysis, and longitudinal tracking | ✅ |
nsd-skill |
Natural Scenes Dataset BIDS validation, task-fMRI processing, and COCO stimulus extraction | ✅ |
nifd-skill |
NIFD dataset BIDS validation, multimodal sMRI/rs-fMRI/dMRI processing for frontotemporal dementia | ✅ |
oasis-skill |
OASIS dataset BIDS validation, sMRI processing, and phenotype extraction for aging/AD research | ✅ |
pnc-skill |
PNC dataset BIDS validation, multimodal sMRI/rs-fMRI/task-fMRI/dMRI processing for developmental studies | ✅ |
ppmi-skill |
PPMI dataset BIDS validation, multimodal sMRI/rs-fMRI/dMRI processing for Parkinson’s disease | ✅ |
rest-mneta-mdd-skill |
REST-meta-MDD multi-site rs-fMRI processing, site harmonization, and depression phenotype extraction | ✅ |
scan-skill |
SCAN/NACC access planning, approved-export staging, phenotype linkage, and multimodal MRI/PET processing | ✅ |
seed-iv-skill |
SEED-IV EEG emotion recognition (4 emotions), feature extraction, and classification | ✅ |
seed-vig-skill |
SEED-VIG EEG vigilance/fatigue detection, feature extraction, and drowsiness classification | ✅ |
tcp-skill |
Transdiagnostic Connectome Project BIDS validation, multimodal sMRI/rs-fMRI/dMRI processing | ✅ |
ucla-cnp-skill |
UCLA CNP BIDS validation, multimodal sMRI/task-fMRI/dMRI processing, multi-disorder phenotyping | ✅ |
ukb-skill |
UKB brain imaging automated processing workflow | ✅ |
Modality
| Skill | Function | Status |
|---|---|---|
eeg-skill |
EEG preprocessing and feature extraction workflows | ✅ |
fmri-skill |
Functional MRI preprocessing and analysis workflows | ✅ |
smri-skill |
Structural MRI preprocessing and analysis workflows | ✅ |
dwi-skill |
Diffusion MRI preprocessing and analysis workflows | ✅ |
pet-skill |
PET imaging workflows (SUVR computation, reference regions, PVC) | ✅ |
asl-skill |
ASL perfusion MRI workflows (CBF quantification, Buxton model) | ✅ |
meg-skill |
MEG processing workflows (source localization, time-frequency, connectivity) | ✅ |
Legend: ✅ Implemented | 🏗️ In Development | ⏳ Planned
🙏 Acknowledgments
Thanks to:
- OpenClaw
- Hermes
- Claude Code
- Karcen/rs-fMRI-Pipeline-Tutorial
- nature-skills
- Open-source neuroscience tools community (MNE-Python, FreeSurfer, FSL, etc.)
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add cuhk-aim-group/neurodiscovery