Hiring research with Thirdwatch MCP: a sourced recruiter brief

A recruiter brief needs evidence for where roles are open, what requirements are disclosed and which details are missing. Thirdwatch's search_jobs searches supported job sources and returns normalized listings. It does not automatically apply to jobs, contact candidates or establish a company's hiring growth rate.

Choose a question that a listing can answer

Start with one role and one geography. For example: which sources have senior Python roles in Bangalore, and what requirements do those listings disclose? Separate this from a different question such as who is available to hire: an open role describes an employer's demand, not a person's availability.

Connect Thirdwatch MCP, then preview search_jobs using estimate_cost. Specify country and tier where applicable. The lite tier limits source fan-out; full uses more applicable sources. Asking for more sources can increase credits and waiting time. An explicit source subset is useful when testing a particular board.

A bounded job search

{
  "query": "senior Python engineer",
  "location": "Bangalore",
  "country": "IN",
  "tier": "lite",
  "max_results": 10
}

Ask your assistant:

Use search_jobs with these inputs. Group the returned jobs by employer. Include title, location, posted date when present, disclosed salary with its currency and the original job URL. Keep missing dates and salaries blank. Summarize repeated skills from the returned descriptions and flag source warnings.

When displaying a result from Remote OK, credit the source exactly as Remote OK and link to the original job URL. Keep that attribution in any exported table or brief.

These are illustrative inputs. They are not a live shortlist or a guarantee of ten results. The returned sample may omit a board, contain duplicates not caught by normalization or include listings that have changed since publication. Open the original job before making a decision.

Read filters carefully

remote_only filters the fetched results; it cannot prove a job permits working from every country. salary_min removes listings with no disclosed salary and compares the source value. Do not use it to compare salaries across currencies. posted_within_days can retain undated jobs, so request a separate undated section instead of claiming every result is recent.

For a company career-site search, use its career-page URL as query and set sources to ["career_sites"]. The explicit source selection avoids sending that URL as a search phrase to other job boards. The Greenhouse scraping tutorial remains available for developers who need the standalone Apify Actor and its different input options.

Add public profile research as a separate step

search_candidates searches public profiles using role and optional filters. get_company_employees accepts a LinkedIn company URL for employer-specific profile research. Neither establishes consent to contact, complete employee coverage or whether someone wants a new role. Do not infer personal email addresses or manufacture a candidate's job preferences.

Ask for a second credit preview before adding these tools. Keep the profile research in a separate section so a reader can distinguish actual job listings from public-profile matches.

Deliver a brief that can be checked

A useful output has a source-backed table, a short interpretation of the returned sample and a gap log. Include the query, geography, tool, selected tier and date of the request. Avoid saying a company is ramping up hiring from one snapshot; that needs comparable repeated observations.

See the hiring research workflow. Thirdwatch MCP spends Thirdwatch credits; running the standalone Actors uses separate Apify billing.