Core
Python >= 3.10 and Git are required.
NeuroClaw
Get Started
Install NeuroClaw, launch the agent, and verify the setup from one place.
Python >= 3.10 and Git are required.
Conda or Mamba for isolation, and nvidia-smi or nvcc for GPU support.
Recommended packages include pypdf, python-docx, openpyxl, and python-pptx.
NeuroClaw provides a setup wizard for runtime configuration, toolchain paths, and Web UI dependencies.
Installation
The bundled installer walks you through Python runtime selection, CUDA/GPU configuration, neuroscience toolchain paths, LLM backend selection, default directories, and Web UI dependencies.
python installer/setup.py--non-interactive for a quick auto-detected setup.git clone https://github.com/CUHK-AIM-Group/NeuroClaw.git
cd NeuroClaw
python installer/setup.py
python installer/setup.py --non-interactive
pip install "fastapi[standard]" uvicorn pypdf python-docx openpyxl python-pptx
Quick Use
You can run NeuroClaw either in the terminal REPL or in the browser-based Web UI.
python core/agent/main.pypython core/agent/main.py --webhttp://localhost:7080 in your browser.--port and --host for remote access or custom binding.python core/agent/main.py
python core/agent/main.py --web
python core/agent/main.py --web --port 8080 --host 0.0.0.0
Install NeuroClaw as a router skill or rule pack so the host agent can enter NeuroClaw mode with NeuroClaw catalogs, NeuroOracle, NeuroBench, model workflows, and neuroimaging skills.
| Host agent | Install command | Generated integration |
|---|---|---|
| Codex | python installer/install_agent_integration.py --target codex |
~/.codex/skills/neuroclaw/ |
| Claude Code | python installer/install_agent_integration.py --target claude-code |
~/.claude/skills/neuroclaw/ |
| Cursor | python installer/install_agent_integration.py --target cursor --scope project |
.cursor/rules/neuroclaw.mdc plus generated references |
| OpenClaw | python installer/install_agent_integration.py --target openclaw |
~/.openclaw/skills/neuroclaw/ |
| Hermes | python installer/install_agent_integration.py --target hermes |
~/.hermes/skills/neuroclaw/ |
| WorkBuddy | python installer/install_agent_integration.py --target workbuddy |
~/.workbuddy/skills/neuroclaw/ |
| QClaw | python installer/install_agent_integration.py --target qclaw |
~/.qclaw/skills/neuroclaw/ |
python installer/install_agent_integration.py --target both
python installer/install_agent_integration.py --target all
python installer/install_agent_integration.py --target all --export ./dist/agent-integrations
After installation, ask the host agent to "use NeuroClaw" or "enter NeuroClaw mode" for neuroimaging and autoresearch work. If the host agent uses a different import location, export the generated skill pack and import or copy it manually.
To let Codex, Claude Code, Cursor, or another host agent use its own built-in model instead of a NeuroClaw LLM API key for autoresearch, start the file-based host-agent loop:
python -m neurooracle.src.hypothesis_cli host-agent-init case1_transdiagnostic --output-dir neurooracle/data/host_agent_runs/case1 --max-rounds 5
The host agent reads tasks/round_XXX_task.json, acts as
NeuroClaw's hypothesis generator, critic, and experiment supervisor,
writes host_outputs/round_XXX_result.json, and advances
the loop with:
python -m neurooracle.src.hypothesis_cli host-agent-next --run-dir neurooracle/data/host_agent_runs/case1
Use these checks to confirm the environment file and skill registry are working as expected.
# Check that the environment file is valid
python installer/setup.py --check
# List registered neuroscience skills (Python)
python -c "
from core.skill_loader.loader import SkillLoader
from pathlib import Path
skills = SkillLoader(Path('skills')).load_all()
for s in skills:
print(s['name'])
"
Once NeuroClaw is running, you can explore the benchmark, FAQ, and homepage pages directly from the site.