The words we use, defined.

Short definitions for the terms that show up across the API, the dataset and the docs. Linked from wherever they first appear.

Guest endpoints: The pages and JSON endpoints LinkedIn serves to a visitor who is not logged in — public job postings above all. LinkFetch's jobs index is built only from these, which is why it needs no account, no session and no extension.Jobs API →
User-as-principal: LinkFetch's model for anything behind a login: profile, company, post, group and search reads run in the end user's own LinkedIn session through the Chrome extension. The user is the principal; LinkFetch never operates a fake or rented account on anyone's behalf.Compliance →
Provenance stamp: The meta.provenance block on every response: source (where the row came from), fetched_at (when), and freshness_days (how old it is). It lets an agent cite what it learned and decide whether to re-fetch.Provenance →
Credit: LinkFetch's billing unit. Each endpoint has a fixed credit cost per request regardless of rows returned, and empty results cost zero. The free tier includes $5 of credit.Pricing →
first_seen_at / last_seen_at: Lifecycle timestamps on every job row: when LinkFetch's crawler first indexed the posting, and the most recent crawl that still found it live. Together they give a posting's observed lifetime independent of the date LinkedIn displays.
is_active: A boolean re-checked on every recrawl. False once LinkedIn stops serving the posting. Filtering on it keeps an agent's index free of expired roles.
Applicant count: The number of applicants LinkedIn displayed for a posting at the last observation, with applicant_count_is_capped set when LinkedIn showed a ceiling such as "over 200" instead of a number. Present on ~8% of rows.Dataset schema →
Applicant velocity: How fast a posting's applicant count is rising, derived from job_applicant_history — the table that records a timestamped applicant count each time a posting is re-observed. A high velocity means a competitive role; a flat series on an old posting often means it is effectively filled.Jobs dataset →
Parsed salary: salary_min, salary_max, salary_currency and salary_interval, extracted from the posting's own salary text so they can be compared as numbers. Present on ~10% of rows; the raw salary_range string is kept alongside.
Geo ID: LinkedIn's numeric identifier for a place. Resolving a city or region name to its geo ID via the locations endpoint gives exact matching instead of free-text substring matching on the location field.Locations API →
Snapshot: The one-time jobs dataset purchase: the whole archive built fresh on the day of purchase as a PostgreSQL dump, CSV and JSONL, with a manifest stating exact row counts and the coverage window. $199.
Live access: The subscription form of the jobs dataset: a read-only Postgres role on a replica refreshed nightly, with unlimited queries and no credits. $49/month.
MCP server: A Model Context Protocol server exposes an API as tools an AI client can call natively. LinkFetch's is the linkfetch-mcp package; it runs locally over stdio and is metered exactly like the REST API.MCP server →
extension_required: The 422 response returned when a session-backed record is not in LinkFetch's cache yet. The Chrome extension captures the page from the user's signed-in tab, and the next call returns the row.Chrome extension →
Envelope: The response shape shared by every endpoint: data holding the typed record or records, and meta holding request_id, credits_charged, rate_limit and the provenance stamp. Errors use the same envelope with a top-level error object.Errors →