Finder

A pattern engine your agent can call.

Finder searches tables for drivers, anomalies, and segment interactions before the report gets written. Use it directly or start with the bundled CLI.

See example output
finder run
finder connect cms_medicaid.csv
rows: 227,000,000 · mode: local run
finder discover --target spending_growth
testing service category, provider, geography, year, price, volume
top pattern: HCBS volume growth explains 43% of increase
strongest segment: agency-based care delivery
report artifact ready for agent synthesis

Inputs

One table, local files, dbt context, warehouse exports, or notes your agent can read.

Run

Statistical sweeps rank drivers, anomalies, and segment interactions worth investigating.

Outputs

Structured findings, caveats, open questions, and artifacts the agent can turn into a report.

What Finder does

It does the search work.

Agents are good at queries and summaries. Finder handles the combinatorial part: testing plausible cuts, ranking useful ones, and returning evidence the agent can use.

Driver discovery

Identify which dimensions, measures, and changes explain the movement in a KPI.

Segment interactions

Search for combinations that disappear in univariate summaries and dashboard rollups.

Anomaly exploration

Flag changes worth investigating and preserve the caveats your agent needs for follow-up work.

Report-ready evidence

Return scale, effect size, context, open questions, and source-backed findings.

Where it runs

Portable by default. Bundled when useful.

Use Finder directly

Call the standalone package from your coding tool, agent harness, or local workflow.

Use the bundled CLI

Start with the Unsupervised CLI when you want Claude Code plus Finder, DeepWork, and analysis defaults.

Add DeepWork when the workflow matters

Pair Finder with reusable Agent Operating Procedures for quality gates and repeatable runs.

Free local use

Free for serious local work. Paid when teams need scale.

Free local CLI preview for agent-led analysis.

Paid software adds larger workloads, managed runs, and governed team use.

Team governance, run history, and shared controls sit in the paid layer.

Enterprise deployment support when data cannot leave your environment.