start
Moving from Fantastic.jobs
Their `_advanced` boolean expressions work here as written, and their parameter names are accepted as aliases. This page lists every difference, side by side.
1What you change, and what you do not
Change the host, the path and the key. Keep the boolean expressions: (react | vue) & (node | golang) & !aws means the same thing on both APIs.
In depthWhy it exists, what it is not, what people get wrong
- Host and path. Their job routes are
/v1/active-ats,/v1/active-jband theexpired-andmodified-variants. Ours is one route,GET https://api.jobopportunitiesapi.org/v1/jobs, for every source, withGET /v1/jobs/closedfor postings that came off their source andGET /v1/changesfor the delta feed. - Key.
Authorization: Bearer <key>. See authentication. - Boolean notation.
&|!:*<-><N>, single-quoted phrases and parentheses are part of our grammar, with the same precedence (!before&before|). The full grammar: the query grammar. - Parameter names.
title_advanced,description_advancedandorganization_advancedare accepted as aliases (the table below).location_advancedis not: places are structured filters here.
2The parameter map
Each Fantastic.jobs name, the JOA parameter it maps to, and the difference if there is one.
In depthWhy it exists, what it is not, what people get wrong
| Fantastic.jobs | JOA | Difference |
|---|---|---|
title_advanced (also advanced_title_filter, title_filter) | title | None in meaning: a boolean search of the job title. |
title | title | Their bare title is Google-style (space is AND, quotes, leading -). Ours takes the same and the full grammar. |
description_advanced (also advanced_description_filter, description_filter) | description_contains | Theirs searches the title and the description as one text. Ours searches the advert text only, and live postings only. Adverts almost always repeat the title, but a term that is only in the title is not matched: add title= for that. |
description | q or description_contains | Not an alias, because it is ambiguous here: q searches title, company and location; description_contains the advert text. |
organization_advanced (also advanced_organization_filter) | company_name_contains | None in meaning: a boolean search of the employer name. |
organization, exclude_organization | company_name_contains or company | Refused with a 400 naming the JOA parameter. Exclude with company_name_contains=!(Amazon | Meta). |
location_advanced, location | country, city, state, region | location_advanced is refused with a 400 naming these. Our places are normalised fields (ISO country codes), not a text to search: a country name is present in the location text of a minority of the rows of that country, so a text search would silently miss most of them. |
date_posted_gte, date_posted_lt | posted_after, posted_before | Not aliases. listed_after is the time a row reached the API, the one to poll on. |
limit, offset | limit, cursor | limit is 1 to 200. Paging is keyset: follow next_via; offset is refused. |
3Their documented examples, side by side
Every example from the Fantastic.jobs advanced searching guide, and the JOA request that answers it.
In depthWhy it exists, what it is not, what people get wrong
| Fantastic.jobs | JOA |
|---|---|
title_advanced=(AI | 'machine learning' | robotics) & !marketing | title=(AI | 'machine learning' | robotics) & !marketing |
title_advanced=senior & (python | rust) & !staff | title=senior & (python | rust) & !staff |
title_advanced=project <-> manag:* | title=project <-> manag:* |
title_advanced='data engineer' | 'data scientist' | 'analytics engineer' | title='data engineer' | 'data scientist' | 'analytics engineer' |
description_advanced='remote work' & (equity | 'stock options') & !contractor | description_contains='remote work' & (equity | 'stock options') & !contractor |
description_advanced=(kubernetes | k8s) & terraform & aws | description_contains=(kubernetes | k8s) & terraform & aws |
location_advanced='United States' & !California | country=US, then drop rows with state CA on your side, or list the states you want in state= |
location_advanced=Germany & !(Berlin | Munich) | country=DE, then drop the two cities on your side, or list the cities you want in city= |
location_advanced='New York' | 'San Francisco' | Boston | city=New York,San Francisco,Boston |
organization_advanced=University & !Harvard | company_name_contains=University & !Harvard |
organization_advanced=(Microsoft | Google | Apple) | company_name_contains=(Microsoft | Google | Apple) |
organization_advanced=OpenAI | 'Open AI' | company_name_contains=OpenAI | 'Open AI' |
curl -s -H "Authorization: Bearer $JOA_KEY" --get \
https://api.jobopportunitiesapi.org/v1/jobs \
--data-urlencode 'title=(engineer | developer) & !manager' \
--data-urlencode 'country=US' \
--data-urlencode 'company_name_contains=!(Amazon | Meta | Google)' \
--data-urlencode "description_contains=remote & (equity | 'stock options')" \
--data-urlencode 'limit=5'4Where the grammars differ
Ours accepts everything theirs does, plus a few forms theirs refuses. Errors are 422 here, 400 there.
In depthWhy it exists, what it is not, what people get wrong
- A space is AND here.
title_advanced=software engineeris a 400 on Fantastic.jobs, which requires an explicit operator between terms. Here it issoftware & engineer. - Word operators too.
AND,ORandNOT(upper case) work as well as&|!, and-wordexcludes. Quote"OR"to search for the word. - Prefix.
manag:*andmanag*are the same. A prefix needs 3 or more characters. A bare word is always exact, as on Fantastic.jobs. - Phrases.
'machine learning'and"machine learning"are both the words adjacent and in order. - Errors. A malformed expression is a 422
query_syntax(Fantastic.jobs answers 400), and the message names the character position. A boolean expression of more than 32 terms or parentheses nested more than 8 deep is a 422query_too_complex. Values may be up to 500 bytes. A refused request costs no records. - Matching. Case-insensitive and accent-insensitive.
OpenAIandOpen AIare different words here too: list both.
This page was rendered 17 November 2026, 07:47 UTC. Every figure on it comes from the endpoint named beside it, and every published request is re-sent against the live API before this site is allowed to build. If something here is wrong, the thumbs-down above reaches a person.