getedgehq/rocketlist

rocketlist 是做什么的？

Read a CV once, work out what the person can actually do, then find the live postings that match, including the ones filed under a title they have never typed into a search box.

The surface you are working with

Rocketlist has no public JSON API. Its robots.txt disallows /api/ for every agent, so this skill works over the public web pages, which need no account and no credential.

• Search: https://rocketlist.ai/jobs?q=<terms>, plus city, seniority, category, industry, stage, founded, funding, investor, sort, page. The result list renders in the browser, so fetch this URL with something that executes JavaScript. A plain GET of a filtered search returns an empty list, not an error.
• Job detail: https://rocketlist.ai/jobs/<slug>. Server rendered, so a plain GET works. Carries the published salary, the required and nice-to-have skills, and job_url, the direct link to the employer's own ATS posting.
• Job sitemaps: about 91,000 live job URLs whose slugs contain the company and the title. A free offline index for testing whether a title exists before searching it.

Exact parameters, taxonomies, parsing recipes and gotchas: references/search-surface.md.

Rocketlist also documents a public, token-free MCP connector at https://rocketlist.ai/mcp with search_jobs, get_job, search_companies and get_company. It was returning HTTP 502 on 2026-09-15. Send one initialize call; if it answers, use it, because its search_jobs takes remote status and skills that the URL cannot express. On any non-200, fall back to the pages and do not retry.

Workflow

1. Read the CV into capabilities. Verbs, objects, counterparties, tools, domain, scope, and the constraints the person stated: location, remote, seniority, salary floor, industries they will not go back to. Ask only for a constraint you genuinely need.

2. Expand into fifteen to twenty-five candidate titles along the five axes in references/title-expansion.md. Keep both the titles the person would have searched and the ones they would not, tagged so you can separate them later. Record the CV evidence behind each one.

3. Kill the invented titles before you spend a search on them.

   python3 scripts/rocketlist_job.py scan --pattern "forward deployed" --pattern "deployment strategist"
   

   This greps the job sitemaps and prints a live posting count per title. Zero means the title does not exist on this board. Drop it silently.

4. Search each surviving title separately. One title per q, widest first. q is free text over the description as well as the title, so long compound queries collapse to nothing. Add city, seniority or investor only when a bare title returns too much. Remote status is not a URL parameter; read it off each card or put remote in q.

   Capture from each result card: title, company, location, salary range, one-line summary, tech stack, and the /jobs/<id> link. Ignore "Match score locked", there is no score for a signed-out visitor.

5. Open the detail page for every role you intend to shortlist.

   python3 scripts/rocketlist_job.py job <slug-or-url>
   

   This prints the structured record: job_salary_range, job_required_skills, job_nice_to_have_skills, job_experience_required, job_seniority, job_location_type, company stage, investors, and job_url. Check the requirements against the CV here. A title that sounded right and asks for eight years of a skill the person does not have is not a fit; cut it.

6. Rank. Dedupe on job_url, since the same role can surface under several titles. Then order by how much of the CV the posting's required skills are covered by, whether the stated experience band contains the person's, whether the location and remote status satisfy their constraint, and whether the salary clears their floor. Do not weight by how exciting the company is.

7. Present the shortlist. Eight to fifteen roles, in two labelled groups:

   • Roles you would have found yourself, the titles matching their current one.
   • Roles you would not have searched for, the expansion. This is the part that earns the skill. For each of these, name the title, and say in one sentence which part of their CV maps onto it.

   One row per role:

   Field	Content
   Title and company	plus company stage and lead investors when notable
   Location	Remote, Hybrid, On-site, or the city
   Salary	the published range verbatim, or "not published"
   Why you fit	the specific CV evidence, not adjectives
   Gap	the requirement they do not meet, stated plainly
   Apply	the job_url from the record, the employer's own posting

   Close with the titles you tested that turned out to be dead ends and why, so they can steer the next pass.

Boundaries

• Rocketlist aggregates public postings; it is not the employer and the listing can be stale. Send the person to job_url and tell them to confirm the details there.
• Salary is published only when the employer published it. Many records read "Not specified". Write "not published" and never estimate a range, infer one from a similar role, or present a market average as this role's salary.
• Stay on the public pages. Do not call /api/, do not create an account, do not sign in, and do not use a signed-in or admin surface even if one is available to you.
• Every role in the output must be a posting you actually retrieved. Never fill a thin shortlist with a plausible role, a company you assume is hiring, or a title you did not verify against the index.
• The site's own counters, currently about 98,800 active jobs across about 4,860 companies, are Rocketlist's numbers. Quote them as such or leave them out.
• A "Gap" line is not optional. A shortlist with no gaps anywhere is a shortlist that was not checked against the postings.