NEO on VS Code
Quick TL;DR
NEO is a local-first AI engineering agent for code, data, and ML workflows.
The VS Code extension lets you run autonomous tasks, debug environments, and manage data locally without uploading your code.
Install the extension, open a project folder, and start interacting with NEO from the sidebar or terminal in under 5 minutes.
What is the NEO Extension?
Local-First Execution
All code and data stay on your machine; no external uploads required
Encrypted Vault
API keys and credentials stored locally, encrypted at rest
Autonomous Workflows
Automatically installs dependencies, handles errors, and self-corrects code
Architecture
Local agent orchestrating code execution, LLM prompts, and environment fixes
Access external services like AWS S3, W&B, Hugging Face if configured
Platform vs VS Code Extension
| Platform | VS Code Extension | |
|---|---|---|
| Setup | No setup—just upload | Install extension, open a project folder, start NEO |
| Data | 50MB files or cloud | Unlimited local files, direct access to project directories |
| Security | Upload to cloud | All operations local, credentials stay encrypted on device |
| Best For | Quick prototyping | Local development, Git workflows, data science projects |
Provider Integrations
AWS S3
Load datasets and model checkpoints locally; configure API keys in vault
Weights & Biases
Track experiments, logs, and artifacts automatically from VS Code
Hugging Face
Access model hub locally; pull/push models securely
Kaggle
Download datasets and competition files directly into project workspace
Multi-Workspace Support
Isolated Contexts
Each workspace runs a separate NEO instance, preventing interference
No Context Leakage
Credentials, secrets, and project state stay workspace-specific
Parallel Execution
Run multiple workspaces and tasks simultaneously without collisions
Use Cases
Data Pipelines
Automate fetching from S3, validating CSVs, and loading into DBs
Experiment Tracking
Train ML models and log metrics to W&B automatically
Environment Fixes
Detect and fix dependency, Python version, and CUDA issues in projects
Model Deployment
Package models and push to cloud registries or local containers
Security
Zero-Knowledge
Code, data, and credentials never leave your machine
Training Opt-Out
Codebase is never used for AI model training or analytics
Full Control
Interrupt, review, and audit all automated actions