vesper-wizard

Zero-friction setup wizard for Vesper — your local MCP-native dataset intelligence layer.

Install

npx vesper-wizard@latest

That's it. The wizard handles everything:

1. Creates ~/.vesper/ directories and local API key
2. Initializes a local credentials vault in unified-key mode (no external API keys required)
3. Installs @vespermcp/mcp-server and auto-configures MCP for all detected agents (Claude, Cursor, VS Code, Codex, Gemini CLI)
4. Verifies the installation
5. Opens browser onboarding during wizard auth (can be disabled with VESPER_WIZARD_AUTO_OPEN=0)

What you get

After the wizard finishes, your AI assistant can immediately use Vesper tools:

| Tool | Description | |------|-------------| | vesper_search | Search 16,000+ datasets via natural language | | discover_datasets | Discover from HuggingFace, Kaggle, OpenML, data.world | | download_dataset | Download any dataset to local storage | | prepare_dataset | Full pipeline: analyze → clean → split → export | | analyze_quality | Deep quality analysis with recommendations | | export_dataset | Export to parquet, csv, feather, jsonl, arrow | | fuse_datasets | Combine multiple datasets with quality checks |

Security

• Local-only: Uses one local key in ~/.vesper/config.toml
• Keyring-backed: Uses OS keyring when available, falls back to local TOML
• No cloud: Zero external API calls during setup
• No external keys: No HuggingFace/Kaggle/Nia key prompts during setup

Config file

The wizard generates ~/.vesper/config.toml:

api_key = "vesper_sk_local_..."
auth_mode = "local_unified"

Post-setup

Restart your IDE and try in your AI assistant:

vesper_search(query="sentiment analysis")
prepare_dataset(query="image classification cats dogs")
analyze_quality(dataset_id="imdb")

License

MIT