API & MCP

search_roles

Search open roles in Australia. Defaults to the data, analytics and ML slice.

Arguments

• qstring

  Words that must all appear in the title, the employer or the role’s vocabulary (kw). Case-insensitive and matches parts of words. Very common words are not in the vocabulary, so python matches on titles only; to search by skill, use match_resume.

• scopedata | all

  data (default) is the data, analytics and ML slice. all is every open role in Australia.

• citystring

  Exact match on the employer’s city, e.g. Sydney. Rows with only a location_raw never match.

• companystring

  Any part of the employer’s name, case-insensitive.

• atsstring

  One of greenhouse, workday, lever, ashby, smartrecruiters, oracle, eightfold.

• remotestring

  Exact match on the employer’s own field, e.g. remote or hybrid. Often unknown.

• sincedate

  Posted on or after this date, YYYY-MM-DD.

• untildate

  Posted strictly before this date, YYYY-MM-DD. With either date set, rows with no date are left out.

• has_salaryboolean

  Only roles publishing a salary band: under 1% of them.

• sortnewest | salary

  newest by default. salary sorts by the top of the band, with unbanded roles last.

• limitinteger

  1 to 200. Default 25.

• offsetinteger

  For paging. Default 0. total says how far there is to go.

• total counts every match, not just this page.
• location is the city, or the raw location where no city could be read. posted falls back to when Muster first saw the role.
• salary_min, salary_max and salary_currency appear only when the employer published a band. Usually they are absent.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "search_roles",
    "arguments": {
      "q": "dbt", "city": "Melbourne", "limit": 1
    }
  }
}

{
  "total": 4,
  "scope": "open data/analytics/ML roles in Australia",
  "offset": 0,
  "results": [
    {
      "id": "smartrecruiters:carsales:744000149506190",
      "title": "Senior Data Engineer - Data Platform & Engineering",
      "company": "carsales",
      "location": "Melbourne",
      "remote": "hybrid",
      "posted": "2026-09-15",
      "apply_url": "https://jobs.smartrecruiters.com/carsales/744000149506190",
      "ats": "smartrecruiters"
    }
  ],
  "snapshot": {"...": "as above"}
}

get_role

One role, by the id that search_roles returns.

Arguments

• idstringrequired

  <ats>:<board_token>:<external_id>

Returns the same fields as a search result, plus four more:

• is_data_role: whether the title passes the data-slice rule.
• skills: the closed vocabulary the résumé matcher scores against (sk).
• vocabulary: the rare words from the ad, which is what q searches (kw).
• identical_openings_at_this_employer: how many open roles that employer’s board has with exactly this title. Muster publishes this instead of a made-up match score. Eight identical openings means a team is being hired, which is worth knowing.

An unknown id is not an error. It comes back as a result that says so, and suggests searching again.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_role",
    "arguments": {
      "id": "smartrecruiters:carsales:744000149506190"
    }
  }
}

{
  "id": "smartrecruiters:carsales:744000149506190",
  "title": "Senior Data Engineer - Data Platform & Engineering",
  "...": "the search_roles fields",
  "is_data_role": true,
  "identical_openings_at_this_employer": 1,
  "skills": ["python", "sql", "snowflake", "postgres", "dbt", "etl", "git"],
  "vocabulary": ["ai-augmented", "anchor", "..."]
}

{
  "error": "no open role with that id in this snapshot",
  "hint": "It may have been filled and left the index, or the id may be from an older export. Search again rather than assuming it is closed."
}

index_health

No arguments. What the index holds and how fresh it is: open counts, counts across the whole index worldwide (index_totals), boards watched, the last sweep, the age of the oldest board, health per ATS, and a caveats array.

caveats states the coverage limits in plain words. Use it to answer questions about how far Muster can be trusted, rather than guessing.

Read boards_failed and each vendor’s incomplete together. A failed board was not fetched at all. An incomplete one was fetched but returned fewer roles than the vendor said it had, and an incomplete board closes nothing until it comes back whole.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {"name": "index_health", "arguments": {}}
}

{
  "open_roles_australia": 9160,
  "open_data_roles_australia": 485,
  "index_totals": {
    "jobs": 334121, "open": 311210, "au_open": 9160,
    "closed": 22911, "boards": 911, "vendors": 7
  },
  "boards_watched": 925,
  "last_sweep": "2026-09-18T22:19:51.092042+00:00",
  "oldest_board_age_hours": 97.32841436,
  "boards_failed": 11,
  "by_ats": [
    {
      "vendor": "workday", "boards": 536, "ok": 500, "failed": 5,
      "incomplete": 31, "fetched": 248068, "new": 24423, "closed": 10304,
      "reopened": 176, "last_run": "2026-09-18T22:19:51.092042+00:00",
      "oldest_run": "2026-09-15T02:18:06.373022+00:00"
    }
  ],
  "roles_with_no_description": 473,
  "caveats": ["...three entries naming the coverage limits..."]
}

match_resume

Ranks open roles against a résumé, with the same scoring the site uses:

• a closed list of skills, each worth more the rarer it is across the index;
• a match on the role title;
• a penalty where the résumé states a level two or more steps below the role’s.

Arguments

• resumestringrequired

  Plain text, up to 60,000 characters.

• scopedata | all

  Default data.

• limitinteger

  1 to 100. Default 25.

• read_from_resume is what the matcher took from the text, so a bad ranking can be traced rather than guessed at. level runs from 0 (graduate) to 5 (head of, director), and is null when the résumé states no level.
• This ranks and never hides. total_ranked is every role in scope and matched is how many scored above zero. A role with nothing in common sinks to the bottom with an empty matched_on; it is not removed. Ties go to the newest.
• title_match is the role title from the résumé found in the job’s title, or null.
• relative_score compares each role with the best match in this response. It is not a probability, and it cannot be compared across calls. Muster has never seen an application, an interview or a hire, so it has no number for those.
• stretch: true marks a role two or more levels above the résumé. It is shown rather than hidden, because a reader may want it.
• unrankable_roles_in_index counts roles with no description to score. They are in the index but cannot rank.
• Text with no known skill or role title gets an error instead of a made-up order.

Sending a résumé to a server

The roles page ranks in your browser: nothing is uploaded. This endpoint is a server, so match_resume always receives the text. It scores it in memory for that one call and does not store or log it, but that is a promise about the server, not something you can check. Sending a request from this page is the same call, so the text goes to the server there too. If the difference matters to you, use the site, or run the ranking yourself from jobs-data.json and the vocabulary in manifest.json.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "match_resume",
    "arguments": {
      "resume": "Analytics engineer, five years. SQL, dbt, Snowflake, Python.",
      "limit": 1
    }
  }
}

{
  "read_from_resume": {
    "skills": ["python", "sql", "snowflake", "dbt"],
    "roles": ["analytics engineer"],
    "level": 3
  },
  "total_ranked": 485,
  "matched": 184,
  "unrankable_roles_in_index": 473,
  "results": [
    {
      "id": "smartrecruiters:carsales:744000149506190",
      "title": "Senior Data Engineer - Data Platform & Engineering",
      "...": "the search_roles fields",
      "matched_on": ["dbt", "snowflake", "sql", "python"],
      "title_match": null,
      "stretch": false,
      "relative_score": 1
    }
  ],
  "note": "Ranked, never filtered. ..."
}

{
  "error": "nothing rankable in that text",
  "hint": "No known skill or role title was found. The vocabulary is a closed list of data/analytics tools and titles; a résumé from another field will not rank, and a made-up ordering would be worse than this error.",
  "vocabulary_size": 72
}