Torre is a professional network with a job board attached, and its job posts carry something most boards do not: the person at the company who published the post, and the visible hiring team. That makes a Torre job dataset partly a dataset about named people, and this guide is mostly about exactly what is collected about them.
Open and closed posts
Torre's search returns closed posts beside open ones. On 1 October 2026, of 2,813 remote “software engineer” posts, 65 were open. Torre listed 16,699 open posts that day out of 363,939 including closed ones. A scraper that does not filter on status delivers an archive and calls it a job feed, so the Torre Jobs & Recruiters Scraper returns open posts by default and closed ones only when status asks for them.
Since 2 October 2026 Torre answers job searches only in the context of a person. The Actor uses a public Torre profile as that context with its skills, experience and preferences removed, so results depend on the filters rather than on someone's profile. Remote search reached 65 of 65 open remote software engineer posts; on-site search works around the main cities of each country and is not exhaustive.
Two populations of posts
58.2% of open posts on 1 October 2026 were aggregated by Torre from other job sites. Those carry the original application link and no company recruiter: they are posted by Torre's own account. The other 41.8% are published by the company on Torre. Every rate is therefore published for both, from 330 open posts measured with the full job page.
| Field | All (330) | By companies (138) | Aggregated (192) |
|---|---|---|---|
| Company domain | 79.1% | 87.7% | 72.9% |
| Salary minimum | 40.9% | 63.8% | 24.5% |
| Person who posted it | 40.6% | 97.1% | 0% |
| Visible hiring team | 69.7% | 99.3% | 48.4% |
| Email in the description | 2.1% | 0% | 3.6% |
An average across the two would describe neither. Anyone after recruiters should read the middle column; anyone after application links, the right one.
What is collected about people
Three levels, each opt-in beyond the first:
- Always, with the full job page: the name, headline and Torre profile link of the person who posted the job and of the visible hiring team, often founders and executives.
- With
includeRecruiterProfile: the poster's public Torre profile — location without coordinates, LinkedIn and other links, summary, current position, languages and top skills. On 90 profiles: location 100%, LinkedIn 77.8%, current position 73.3%. Charged once per person. - With
outputType: recruiters: one row per person, deduplicated, with the jobs they posted.
Things that are never delivered: team members Torre hides on the public job page, screening messages written by candidates (only their count), and any email address outside the company domain that Torre did not publish.
Who is excluded, and how
People whose Torre profile places them in the EU, EEA, UK or Switzerland — 4 of 90 profiles measured — are left out unless the operator turns on includeEuRecruiters, which the input describes as requiring a lawful basis. To apply that rule the Actor reads the public profile of every person it is about to deliver, once per person per run, and does not charge for that read.
In recruiter mode an excluded person is simply absent. In job rows the row keeps its shape and every personal field becomes null, with euRedacted: true. A person whose profile cannot be read or has no country is delivered as location unknown, and the run summary counts them, so the gap in the rule is visible rather than silent. GDPR's information duty for data not obtained from the person, Article 14, is the reason that line sits at the EU border.
Email patterns are guesses
Torre publishes almost no emails: 2.1% of 330 descriptions contained one. With inferEmailPattern, the Actor checks the company domain's MX records by DNS, names the provider, and proposes a format such as first.last. On 1 October 2026 all 45 domains measured had no email evidence, so every pattern was a guess from the provider at confidence 0.10, flagged inferred: true, and delivered free. No email is sent and no mail server is probed, so catchAll is always null. A guessed address is a hypothesis about a format, not a verified contact.
What robots.txt and the terms say
Fetched on 10 October 2026, torre.ai/robots.txt allows /search/jobs exactly, disallows /search/jobs?*, and disallows /api/. The full job page and profile reads use /api/ paths, so they are made against an explicit disallow. The search itself runs on search.torre.co, whose robots.txt returned 404 the same day. Torre's terms list spidering, crawling or scraping the content of the product among prohibited uses.
Those facts are stated here because they are the facts. LinkedIn is the other source in this cluster read against a blanket disallow — what its file says — and the comparison across all five boards is in the cluster overview. For an Actor that drew the line at not reading people at all, see Wellfound's refusal.