Bigml

@bigml list all projects in my account and show me which ones have active correlations or models

Based on the project listing, you have 4 projects: 'Customer Churn Analysis' (3 correlations, 2 models), 'Sales Forecasting Q4' (0 correlations, 5 models), 'Sentiment Analysis Pipeline' (1 correlation, 0 models), and 'Archived_2023' (0 resources). The Customer Churn Analysis project appears most active with recent correlation runs from the past week. Would you like me to retrieve details on any specific project or its resources?

This example demonstrates the MCP's ability to traverse project hierarchies and aggregate resource counts. Requires API key with read permissions across projects. Response time scales with the number of projects in your account—expect 2-5 seconds for accounts with 50+ projects.

@bigml create a new project called 'Product Recommendation Engine' and set up an external connector to our Postgres analytics database

I've created project 'Product Recommendation Engine' (ID: project/abc123def456). Next, I'm setting up the external connector—I'll need your Postgres connection details: hostname, port, database name, and credentials. Once you provide those, I'll create the connector resource and link it to this project so you can pull training data directly from your analytics tables without manual exports.

This write-heavy example shows project creation and external connector setup in sequence. The MCP will create persistent resources in your BigML account—these count toward your subscription limits. External connector credentials are sensitive; ensure your API key has connector creation permissions and review BigML's security practices for stored credentials.

@bigml pull correlation analysis results from the 'Customer Churn Analysis' project and summarize which feature pairs have the strongest relationships

I retrieved 3 correlation analyses from the Customer Churn Analysis project. The strongest relationships are: (1) 'monthly_charges' and 'total_charges' show 0.89 positive correlation—expected since total accumulates monthly; (2) 'contract_type' and 'churn_rate' show -0.62 correlation—month-to-month contracts correlate with higher churn; (3) 'tech_support_calls' and 'churn_rate' show 0.54 correlation—customers needing frequent support are more likely to leave. Consider prioritizing contract incentives and support quality in your retention strategy.

This synthesis example pairs BigML's correlation data with AI reasoning to surface actionable insights. The MCP fetches raw correlation matrices; the AI interprets statistical significance and business implications. Correlation results depend on your dataset's feature engineering—garbage in, garbage out applies.

When BigML's project grouping beats ad-hoc model sprawl

A 5-person data science team runs weekly experiments on customer churn models and needs to keep track of which datasets, transformations, and correlation analyses belong to each sprint. BigML's project tool lets them create a named container for each experiment cycle, then retrieve project metadata to audit what ran when. This works well if your team already uses BigML for model training and wants lightweight organization without standing up MLflow or Weights & Biases. The trade-off: if you need cross-tool lineage tracking or your models live in scikit-learn and PyTorch outside BigML, the project scope is too narrow. Use this MCP when your ML stack is BigML-native and you want a simple grouping layer for 10-50 experiments per quarter.

Connecting BigML to your data warehouse for one-time ingestion

A fintech startup needs to pull transaction data from their Postgres warehouse into BigML for fraud detection modeling, but they don't want to export CSVs manually every week. The external connector tool lets an engineer configure the database link once, then retrieve connector details to verify the connection state before each model run. This is the right call if you have a stable schema in Postgres, Redshift, or MySQL and you're running batch predictions on a schedule. It falls apart if your data sources change frequently or you need real-time streaming ingestion—BigML connectors are designed for periodic pulls, not live pipelines. If your data team is already comfortable with dbt or Airbyte for orchestration, those tools handle connector management better at scale.

When listing correlations helps prioritize feature experiments

A product manager at a SaaS company ran 30 correlation analyses last quarter to understand which user behaviors predict upgrade conversions, and now they need to review which correlations were statistically significant before planning the next A/B test roadmap. The list correlations tool lets them paginate through results and filter by date or strength, surfacing the top candidates without re-running the analysis. This works if your team uses BigML for exploratory stats and you need a lightweight audit trail. The limit: if you're running hundreds of correlations or need to compare results across multiple tools like Amplitude or Mixpanel, BigML's list view gets cumbersome. Use this when your correlation workflow lives entirely in BigML and you're reviewing 10-50 analyses per month.