RocketList

ai.rocketlist由 rocketlist.ai 發布官方發布者v1.0.0更新於 Oct 24, 2026

Search current startup jobs and hiring companies through RocketList's public read-only data.

已驗證Streamable HTTP可網頁執行網頁搜尋與抓取資料與分析

概覽

AI 產生的概覽

以唯讀方式存取 RocketList 的公開新創職缺資料,讓助理搜尋在招職缺與招募公司,並取得彙總的招募統計。

可以做什麼
  • 依自由文字查詢及地點、類別、資歷等篩選條件搜尋公開在招職缺
  • 依 slug 或 UUID 取得單一職缺的公開詳情
  • 搜尋目前至少有一個在招職缺的招募公司
  • 依 UUID 取得某公司的公開資訊及其近期在招職缺
  • 對所有符合條件的職缺做彙總統計:總數、熱門城市、國家、公司、募資階段、工作型態、職缺類別、資歷與公開薪資
適用情境
適合想瀏覽目前新創職缺、比較不同城市、國家或募資階段的招募熱度,或借助助理研究新創招募市場的人。它是探索入口,會導向官方應徵頁面,而非求職申請管理工具。
執行需求
透過 streamable HTTP 連線至 rocketlist.ai 的遠端 MCP 端點;未宣告任何憑證、API 金鑰或環境變數。需要能連線至該端點的網路。
安裝前請注意
資料公開且唯讀,但職缺可能在每日更新之間失效,應以官方應徵頁面為準。涵蓋範圍因雇主、國家與欄位而異,缺漏值不會被補齊,因此彙總統計可能偏低。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 RocketList,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式,或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南,先向你展示發布者、端點和登入方式,經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/ai.rocketlist%2Frocketlist/install.md,並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前,先告訴我它的發布者、端點和登入方式,等我確認。不要把 API 金鑰寫進命令或我會提交的檔案;問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "rocketlist": {
      "type": "http",
      "url": "https://rocketlist.ai/mcp"
    }
  }
}

README

RocketList Startup Jobs

[Daily dataset] [Latest release] [Data license: ODC-BY 1.0]

The open data layer for startup hiring. Updated every day by RocketList.

Live roles from funded startups around the world, normalized into one documented dataset for job seekers, researchers, developers, and AI agents.

109,695 active jobs · 7,297 companies · generated 2026-10-27 09:49:00 UTC

Top locations in this snapshot: Albania: 5 · Algeria: 8 · Argentina: 193 · Australia: 899 · Austria: 131.

Download

The full snapshot lives in the stable latest release, not Git history. SHA-256 checksums, generation metadata, and the complete audit result ship beside every snapshot.

bash
wget https://github.com/rocketlist-ai/startup-jobs/releases/latest/download/jobs.parquet
python
import pandas as pd
jobs = pd.read_parquet(    "https://github.com/rocketlist-ai/startup-jobs/releases/latest/download/jobs.parquet")
berlin_ai = jobs[    jobs["city"].fillna("").str.contains("Berlin", case=False)    & jobs["category"].fillna("").str.contains("AI|Data|Engineering", case=False)]print(berlin_ai[["company_name", "title", "url"]].head(20))

What is included

The jobs dataset contains factual discovery metadata: company, title, normalized role and seniority, location, compensation when explicitly available, skills, canonical application URL, source platform, and first/last-seen timestamps. The companies dataset adds stage, funding, investors, industry, headquarters, and careers URLs where available.

Schemas are versioned in schema/jobs.schema.json and schema/companies.schema.json. A browsable 100-record sample is committed under sample/, while daily aggregate changes live under changes/. See the public methodology and quality checks for provenance, denominators, audit guarantees, and limitations.

What is deliberately excluded

  • Full job descriptions or copied HTML
  • Raw ATS responses and crawler payloads
  • Embeddings, prompts, traces, or enrichment internals
  • Candidate, account, saved-job, application, or matching data
  • RocketList's ranking and recommendation logic

The exporter uses an explicit allowlist and the audit fails if a forbidden field appears.

Use it with agents

Install the user-facing RocketList skill in Claude Code, Codex, Cursor, or another compatible agent:

bash
npx skills add rocketlist-ai/startup-jobs --skill rocketlist-job-search

It searches and filters the current snapshot, handles CV-to-role matching, and links users directly to applications. The bundle is also readable at skills/rocketlist-job-search, so any agent can follow it without an installer.

The accompanying distribution loop explains how the dataset, skill, search pages, and recurring data stories compound into discoverability and traffic.

For bulk analysis, give an agent the Parquet URL and the relevant schema. For lower-latency conversational search, connect the RocketList MCP when available:

text
https://rocketlist.ai/mcp

Its official MCP Registry manifest is versioned in server.json.

Example prompt:

Use the RocketList dataset to find active Series A–C companies in Berlin hiring product managers. Return the canonical application links and explain the filters you applied.

Update model

The workflow runs daily at 04:17 UTC:

  1. Fetch all current public catalog rows through paginated API reads.
  2. Normalize join keys, field types, lists, and stable public IDs.
  3. Retain active, non-duplicate jobs and active or referenced companies.
  4. Cross-foot totals, reconcile API counts, verify unique IDs and URLs, test company references, and reject forbidden fields.
  5. Replace the full assets on the stable latest release.
  6. Commit only small samples, statistics, and daily change summaries.

No credentials are required to reproduce the export:

bash
python -m pip install -r requirements.txtpython scripts/export_dataset.pypython scripts/audit_dataset.py

Accuracy and limitations

RocketList aggregates company career pages and ATS sources. A listed role can close between daily refreshes; the canonical application page is authoritative. Coverage varies by employer, country, and field, and missing values are never imputed for public statistics. See stats/latest.json for field-level denominators.

License and attribution

The dataset is available under ODC-BY 1.0; code is MIT licensed. Attribute RocketList with links to https://rocketlist.ai and this repository. Employer names and trademarks belong to their respective owners, and source postings remain subject to their publishers' terms.

Corrections

Open an issue for a missing company, stale role, broken URL, or schema problem. See CONTRIBUTING.md for the data-safety rules.

來源:README.md,提交 e4b1cc5

工具

5

search_jobs

Search jobs

Search public active Rocketlist jobs. Returns bounded, non-personal job data only. All list arguments are arrays of strings; passing a bare string is a validation error. Filters: - query: free-text over job title, company, location and category. Matched as prefix tokens, so 'engineer' also matches 'Engineering'. - skills: matched against the job's listed tech stack. - domain: exact company industry/vertical, e.g. ['Fintech']. - city: substring of the job's city or country. - remote_only: keep only remote roles. - investor: companies backed by these investors, e.g. ['Index Ventures']. Common aliases resolve ('a16z' finds Andreessen Horowitz). Populated for about 83% of hiring companies, so a company with no recorded investors is never returned when this is set. - stage: exact company funding stage from ['Pre-Seed', 'Seed', 'Series A', 'Series B', 'Series C', 'Series D', 'Series D+', 'Series E', 'Series F', 'Series G', 'Growth', 'Public', 'Acquired', 'Bootstrapped'], case-insensitive. About 12% of hiring companies have no stage recorded and are excluded when this is set. - min_experience_years / max_experience_years: bounds on the years of experience the role requires at minimum. max_experience_years=5 means 'requires 5 years or fewer'. Only about 56% of jobs state a requirement; the rest are excluded when either bound is set. - yc_batch: Y Combinator batches, e.g. ['F26'] or ['Fall 2026'] (W/X/S/F + year). Covers the batches in the reviewed cohort list (currently YC Fall 2026). - founder_origin: countries the founders come from (where they grew up or studied), by name or ISO code, e.g. ['Germany'] or ['DE']. Company-level, reviewed for the same cohorts as yc_batch; a company matches if any founder does. - near: a place, as an address, a street corner ('3rd and Folsom, San Francisco') or 'lat,lng'. Keeps companies whose office (the reviewed street address on getedge.cc/map) is within radius_m metres (default 1000) or walk_minutes on foot (64 m of straight line per m

輸入參數citynearlimitquerystagedomainskillsinvestorradius_myc_batchremote_onlywalk_minutes另有 3 個

get_job

Get job details

Get public details for one active Rocketlist job by slug or UUID.

輸入參數job_id必填

search_companies

Search hiring companies

Search public Rocketlist hiring companies. Returns bounded, non-personal company data only. Only companies with at least one active job are returned, matching the company list on rocketlist.ai. All list arguments are arrays of strings; passing a bare string is a validation error. Filters: - query: substring of the company name or slug, or a whole sector word matching the company's industry, vertical or subvertical. 'fintech' finds companies whose industry is FinTech; 'acme' finds Acme Robotics by name. Sector matching is by whole word, so 'ai' matches the AI & ML vertical and not Blockchain. - investor: companies backed by these investors, e.g. ['Index Ventures']. Common aliases resolve ('a16z' finds Andreessen Horowitz). Populated for about 83% of hiring companies, so a company with no recorded investors is never returned when this is set. - stage: exact company funding stage from ['Pre-Seed', 'Seed', 'Series A', 'Series B', 'Series C', 'Series D', 'Series D+', 'Series E', 'Series F', 'Series G', 'Growth', 'Public', 'Acquired', 'Bootstrapped'], case-insensitive. About 12% of hiring companies have no stage recorded and are excluded when this is set. - yc_batch: Y Combinator batches, e.g. ['F26'] or ['Fall 2026'] (W/X/S/F + year). Covers the batches in the reviewed cohort list (currently YC Fall 2026). - founder_origin: countries the founders come from (where they grew up or studied), by name or ISO code, e.g. ['Germany'] or ['DE']. Company-level, reviewed for the same cohorts as yc_batch; a company matches if any founder does. - near: a place, as an address, a street corner ('3rd and Folsom, San Francisco') or 'lat,lng'. Keeps companies whose office (the reviewed street address on getedge.cc/map) is within radius_m metres (default 1000) or walk_minutes on foot (64 m of straight line per minute). Results carry office_address, distance_m and walk_minutes, nearest first. - radius_m / walk_minutes: size of the near area; give one.

輸入參數nearlimitquerystageinvestorradius_myc_batchwalk_minutesfounder_origin

get_company

Get company details

Get public details and recent active jobs for one hiring company by UUID.

輸入參數company_id必填

hiring_stats

Hiring statistics

Aggregate counts over ALL matching public Rocketlist jobs, not a sample: total jobs and companies, the top values by city, country, company, domain (vertical), company stage, work mode (remote/hybrid/on-site), job category and seniority, and published pay with sample sizes. Use it for 'how many', 'which cities/companies hire most', share and pay questions; use search_jobs for the roles themselves. Questions the Startup Hiring Map can answer (at most one city or country, optionally domain and remote_only) are counted from the same snapshot as getedge.cc/map, so totals match the map exactly (job location; a multi-city role counts in each city). Other filters use the live catalog; the result's `source` says which. Takes the same filters as search_jobs except skills, plus: - group_by: subset of ['city', 'country', 'company', 'domain', 'stage', 'work_mode', 'category', 'seniority'] (default: all). - top: values listed per dimension, 1-50 (default 10). - query: free-text over job title, company, location and category. Matched as prefix tokens, so 'engineer' also matches 'Engineering'. - domain: exact company industry/vertical, e.g. ['Fintech']. - city: substring of the job's city or country. - remote_only: keep only remote roles. - investor: companies backed by these investors, e.g. ['Index Ventures']. Common aliases resolve ('a16z' finds Andreessen Horowitz). Populated for about 83% of hiring companies, so a company with no recorded investors is never returned when this is set. - stage: exact company funding stage from ['Pre-Seed', 'Seed', 'Series A', 'Series B', 'Series C', 'Series D', 'Series D+', 'Series E', 'Series F', 'Series G', 'Growth', 'Public', 'Acquired', 'Bootstrapped'], case-insensitive. About 12% of hiring companies have no stage recorded and are excluded when this is set. - min_experience_years / max_experience_years: bounds on the years of experience the role requires at minimum. max_experience_years=5 means 'requires 5 years or fewer'. Only about 56%

輸入參數topcitynearquerystagedomaingroup_byinvestorradius_myc_batchremote_onlywalk_minutes另有 3 個

版本歷史

1
  1. v1.0.0最新Sep 21, 2026

評分

已滿足適用檢查項的 100%

該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分