Kurobara
Kurobara is an open-source, headless engine for building and enriching B2B lists from the command line or an API. It is designed for humans, scripts, and coding agents that need durable runs, explicit budgets, provider provenance, and machine-readable outputs.
What it does today
- Imports CSV or JSONL datasets.
- Uses imported organization domains directly as the source of a contact
shortlist, with explicit field mapping and deterministic diagnostics.
- Finds companies from country, industry, keyword, and employee-count filters.
- Builds an obfuscated contact shortlist from selected companies.
- Reveals professional identities and resolves work emails for an explicit
selection.
- Verifies work emails as a separate, explicit action.
- Applies versioned enrichment recipes to existing datasets.
- Stores immutable, workspace-scoped GTM Context revisions with explicit
assertion provenance and human-only policy confirmations.
- Compiles versioned, provider-neutral Plays from an exact Context revision,
with bounded previews, budgets, authority gates, and nosend delivery.
- Executes approved Plays through fenced PostgreSQL claims and restart-safe
stage checkpoints, then returns the exact result dataset, materialization, Workbook, settled cost, provider-call count, and export readiness.
- Projects datasets through bounded Workbooks with cell status, provenance,
freshness, confidence, cost, errors, redaction, selection reasons, and append-only human approvals.
- Persists runs, checkpoints, costs, provenance, retries, cancellations, and
ambiguous outcomes in PostgreSQL.
- Exposes the same product logic through REST, a TypeScript SDK, and a