MCP Packs
ML/AI Engineer Pack
Experiment, evaluate, and deploy with model/tooling MCP servers.
ML/AI Engineer Pack
Experiment with models, evaluate results, and deploy with ML-focused MCP integrations. Connect model hubs, data platforms, and ML tooling.
Recommended Servers
| Server | Purpose |
|---|---|
| Hugging Face MCP | Models, datasets, spaces |
| LlamaIndex MCP Toolbox | Database + retrieval utilities |
| Snowflake MCP (Cortex/SQL) | Data + AI from the warehouse |
| GitHub MCP | Experiment tracking in repos |
Example Configuration
{
"mcpServers": {
"huggingface": {
"command": "npx",
"args": ["-y", "@huggingface/mcp-server"],
"env": { "HF_TOKEN": "${HF_TOKEN}" }
},
"llamaindex": {
"command": "npx",
"args": ["-y", "@llamaindex/mcp-toolbox"],
"env": { "OPENAI_API_KEY": "${OPENAI_API_KEY}" }
},
"snowflake": {
"command": "npx",
"args": ["-y", "@snowflake/mcp-server"],
"env": {
"SNOWFLAKE_ACCOUNT": "${SNOWFLAKE_ACCOUNT}",
"SNOWFLAKE_USER": "${SNOWFLAKE_USER}",
"SNOWFLAKE_PASSWORD": "${SNOWFLAKE_PASSWORD}"
}
}
}
}Common Workflows
- Search Hugging Face for models, download datasets, run experiments
- Query data warehouse for training data, build retrieval pipelines
- Track experiments and share results with the team
Safety First
Be mindful of data privacy when querying datasets. Use separate API keys for experimentation vs. production. Monitor costs for compute-heavy operations.