Blog  ›  Search strategy

How AgentFlow AI finds jobs, and what it will not do

The sources, the method, and the deliberate limits. Written so you can judge the results.

Job tools rarely say where their data comes from, which makes their coverage impossible to assess. Here is ours.

Where the jobs come from

Two kinds of source. Public job APIs that exist to be consumed: Remote OK, Remotive, Jobicy, Arbeitnow, Adzuna and Careerjet. And the job board endpoints that applicant tracking systems publish so that company careers pages can render: Greenhouse, Lever, Ashby, Workable, SmartRecruiters, Recruitee and Teamtailor.

On 30 September 2026 that produced 22,182 live postings.

What we will not do

We do not scrape consumer job boards. Not LinkedIn, not Indeed, not the regional boards. Those sites forbid commercial reuse and block automated clients, and building a paid product on them would be both a legal risk and a technical one.

It costs us coverage in some markets, particularly outside technology, and we would rather tell you that than pretend the index is complete.

How matching works

Your CV is read in your browser and never uploaded. From it we derive a field, a level, a skill set and a location preference. Postings are then scored against that profile with plain keyword logic, not a model, which means every result can be explained: the reasons are shown under each role.

Known limits, stated plainly

• Technology fields are better covered than others, because ATS boards skew technical.
• Architecture is our weakest field, and the reason is structural rather than fixable.
• A posting that exists may not appear. We show what our sources publish, not everything.

Why we tell you this

Because a job search runs on judgement, and judgement needs to know what it is looking at. A tool that hides its coverage is asking you to trust a number you cannot check.