# HUNT Workforce Intelligence

HUNT Workforce Intelligence: people flows at 1,000+ public companies (hires, departures, which roles and seniority levels), built from public professional profiles. Company × month, 2010 to the latest release. Use it to answer: who is growing or shrinking, where engineers or executives are leaving, who is building an AI or sales team, where to source candidates for a role, who competes for the same talent.

How to work:
1. Call `guide` once first — it gives signal codes, metric names, role families, grades, sectors, regions and the latest month.
2. Find a ticker by name with `screen` (q=<name>, limit=5). Never guess tickers.
3. Market-wide question → `overview` or `signals`. One company → `company`. One role → `talent`.

How to read the numbers — tell the user this when it matters:
- Every `*_pct` value is the company's percentile among companies of the SAME month (0..1). `*_d12` is the change of that percentile over 12 months. Raw counts are not comparable across months (recent hires are under-observed), so never compare raw levels over time.
- `score` is workforce momentum: mean percentile of net flow, net tech inflow and net inflow of senior/lead/exec grades. 0.5 is the middle of the market.
- Each row is computed on the 36 months strictly before `as_of`. A signal is a slow structural shift, not breaking news.
- Companies with `coverage_tier=low` are not ranked and get no signals.
- This is not investment advice and not a price signal: the only validated link is between early churn and next-year revenue growth; against stock returns it is zero. Employer-to-ticker matching is wrong in roughly 8% of cases — sanity-check surprising results.
- No salary data, no individual people. Only company-level aggregates.


## Connect

- MCP server (Streamable HTTP, no authentication): https://intel.huntshare.tech/mcp
- REST: https://intel.huntshare.tech/v1/... (OpenAPI: https://intel.huntshare.tech/openapi.json)
- Anonymous limits: 60 calls/hour, 300 calls/day per IP, up to 100 rows per call. `guide` is free.

## Tools (MCP name → REST path)

### guide → GET /v1/guide

Call this first, once. Returns signal codes with their meaning, metric names, role families, grades, sectors, regions, sortable columns and the latest available month. Does not count against the rate limit.

Arguments: no arguments.

### overview → GET /v1/overview

The whole market in one call: sector medians, fastest improving and deteriorating companies, number of companies under each signal, sector momentum by year, and the share of each role family in hiring. Start here for market- or sector-level questions.

Arguments: as_of, region.

### signals → GET /v1/signals

Companies whose position shifted sharply over 12 months. Risk: contraction, hiring_freeze, tech_drain, top_drain, exec_exodus, veterans_leaving. Growth: hiring_surge, top_magnet. Shift: ai_build, sales_build. `limit` is per signal. Before drawing conclusions about a company, confirm with `company`.

Arguments: as_of, region, sector, kind, code, limit.

### screen → GET /v1/screen

Companies of one month with percentiles, yearly changes and signals. Filter by region, sector, signal; search by name or ticker with `q`; sort by any column listed in `guide`. This is also how you find a ticker: q=<company name>, limit=5.

Arguments: as_of, region, sector, tier, q, signal, sort, desc, limit.

### company → GET /v1/company/{ticker}

One company by ticker: monthly series of percentiles and signals, the role × grade matrix of hires and departures now and a year ago (who exactly is joining and leaving), tenure of leavers, and peers from the same sector and region. Unknown ticker returns not_found — look it up with `screen`.

Arguments: ticker, as_of, months, peers.

### talent → GET /v1/talent

mode=poach: companies losing this role (more departures than hires) — where to source candidates. mode=hire: companies hiring it most — who you compete with for talent. Counts are observed records over the 36-month window; `*_share_d12` is how the role's share in the company's departures or hires changed over a year.

Arguments: family, grade, mode, as_of, region, sector, limit.

### compare → GET /v1/compare?tickers=A,B

One metric for up to 10 tickers over time: the last `months` months, every `step`-th month. Metric is `score` or any `*_pct` column from `guide`.

Arguments: tickers, metric, months, step.