@brightdata/enrich

Headless CLI for the Open Enrich engine. Enrich any CSV with company and person data using Bright Data's web infrastructure + LLM agents.

Install

npx @brightdata/enrich --version
# or
npm install -g @brightdata/enrich

First-run setup

You'll need a Bright Data account (with two zones: SERP API + Web Unlocker) and an OpenRouter API key.

enrich login

Or set env vars instead:

BRIGHT_DATA_API_KEY=...
BRIGHT_DATA_SERP_ZONE=...
BRIGHT_DATA_UNLOCKER_ZONE=...
OPENROUTER_API_KEY=...

Env vars always take precedence over the stored config.

Usage

# Describe what you want in natural language
enrich leads.csv --describe "company size, recent funding, tech stack"

# Or pick specific preset fields
enrich leads.csv --fields employee_count,industry,funding_stage

# See what would happen without making API calls
enrich leads.csv --describe "..." --dry-run

# Specify the identifier column explicitly
enrich leads.csv --describe "..." --column email_address

# Write to a specific output path
enrich leads.csv --describe "..." --output enriched.csv

# Stream NDJSON events for programmatic use
enrich leads.csv --describe "..." --json | jq .

Commands

| Command | Description | |---|---| | enrich <input.csv> | Enrich the given CSV (the default command). | | enrich login | Interactively configure and store credentials. | | enrich logout | Remove stored credentials. |

Options

| Option | Description | Default | |---|---|---| | -o, --output <path> | Output CSV path. | <input>.enriched.csv | | --column <name> | Identifier column. | auto-detected | | --first-name <name> | Column containing the person's first name. | — | | --last-name <name> | Column containing the person's last name. | — | | --describe <text> | Natural-language description of the fields you want. | — | | --fields <names> | Comma-separated preset field names (alternative to --describe). | — | | --concurrency <n> | Rows enriched in parallel. | 30 | | --no-resume | Overwrite the output instead of auto-resuming. | resume on | | --dry-run | Resolve config and exit without making API calls. | off | | --json | Stream NDJSON events to stdout; do not write a CSV. | off | | -q, --quiet | Suppress progress; print only the final summary. | off | | -v, --verbose | Show every agent_progress event. | off |

Provide either --describe (natural language) or --fields (preset names). Run enrich with no arguments to print full help.

Identifier & name columns

The CLI auto-detects the column that identifies each row (e.g. a company domain, website, or email). Override it with --column when detection guesses wrong. For person-level enrichment, point --first-name/--last-name at the relevant columns so the agents can disambiguate people at the same company.

How resume works

If you re-run with the same --describe/--fields and the output file already exists, the CLI checks the header and skips rows that already have output. Interrupted runs (Ctrl-C) auto-resume on the next run. Pass --no-resume to overwrite.

Cost

Bright Data charges per tool call (~$0.0015). OpenRouter charges per LLM token. Use --dry-run for a rough estimate before kicking off large runs.

License

MIT