kendreaditya/levels-fyi

levels-fyi

Query levels.fyi compensation data, job listings, and company info from the shell. Unofficial but works via their public JSON endpoints.

Subcommands (run levels <cmd> --help for each)

• search-company <name> — find company by name → get ID/slug
• list-companies — all companies (~61K universe)
• company-overview <slug> — ticker, public/private, valuation, HQ, vesting, tags, related companies. Use --full for all fields.
• company-salaries <company> <job-family> — salary distribution for a company + role family. Each sample includes yearsOfExperience so you can compute level-to-YoE mapping.
• jobs — search job listings (filters: company, location, level, etc.)
• job-family <slug> — details for a job family (e.g. software-engineer)
• job-families — all job family categories
• industries — all industries
• locations — all locations by region
• benefits — compare benefits across companies
• leaderboard — top-paying companies

Usage examples

levels search-company "Meta"                         # find Meta's slug
levels company-salaries meta software-engineer       # all SWE comp at Meta
levels jobs --company google --level L6              # L6 jobs at Google
levels leaderboard --job-family software-engineer    # top-paying SWE companies
levels benefits meta google apple                    # compare benefits

All subcommands support --json for machine-readable output (pipe through jq).

When to use

• User asks about compensation at a specific company/level.
• Compare TC across multiple offers.
• Research comp bands before negotiating.
• Pull job listings by level/company/location.

When NOT to use

• Anything requiring login (your saved favorites, my-profile data).
• Real-time salary offers (data is crowd-sourced + lagging).
• Negotiation advice beyond raw numbers — use Blind skill for commentary.

Pairs well with

• blind: for anecdotal comp/negotiation threads (levels = numbers, blind = stories)
• jq: all output supports --json for piping

Gotchas

• Response uses AES-encrypted payloads; pycryptodome is required (auto-installed in config venv).
• Company slugs are normalized (spaces → hyphens, lowercase). Use search-company first if unsure.
• Job family slugs: software-engineer, product-manager, data-scientist, etc. Run job-families to list.
• company-overview returns estimated_valuation and emp_count from levels.fyi's stored snapshot — these can be months to years stale for fast-growing companies (e.g. Anthropic shows emp_count=45, Stripe valuation=$95B). For live valuations use yfinance (public) or external sources (private). The ticker and company_type fields ARE reliable for public/private classification.