Workflows: describe what to watch, and get told when it posts

Most hiring data lands in a dashboard nobody opens. Workflows are the opposite. You describe what you want to watch in one sentence, JobsPipe keeps watching new postings for you, and when a match appears it reaches you where you already work. It is live today for every account at jobspipe.dev/dashboard/workflows.

One sentence in, one workflow out

The builder is a chat, and it understands hiring language. “Alert me when fintech companies in Germany start hiring for Kubernetes” becomes a country filter, a skills filter, an industry filter and an intent, without you touching a form. Behind the chat sits a model (Claude Sonnet 4.5 by default, switchable in the header) with three built-in tools: it can search the live index to check volume, it can scan a company’s website to see what technologies it runs, and it can propose a workflow.

The flow is deliberately short. It asks at most one clarifying question, and only when the request is too vague to configure. It then runs a single search with its candidate filters to see how much volume you would get, and proposes a card. Nothing is created until you click the button on that card. The builder cannot create a workflow on its own, and it will say so if you ask it to.

Two modes: companies and jobs

Every workflow runs in one of two modes, and picking the right one is most of the work.

• Companies fires once per company that starts hiring for what you described. This is the mode for buying signals, target-account lists, competitor tracking and BD leads. A company that opens ten matching roles this week is one alert, not ten.
• Jobs fires for every matching posting. This is the mode for a recruiter sourcing feed, a job-board inventory webhook, or a personal alert where you want each role.

The builder picks the mode from your wording and shows it on the card, so you can change it before anything is created.

The intent sentence is the precision layer

Filters are broad on purpose. Titles and skill slugs catch the candidates; they cannot tell whether a consultancy that mentions Kubernetes is a company adopting Kubernetes. That is what the intent is for. When your request carries meaning beyond the literal filters, the builder writes a one-sentence intent in your own terms, for example “financial services companies adopting Kubernetes in their own technology infrastructure”. Every match is re-judged against that sentence by AI before it alerts you. A plain title feed or a watch on one named company needs no intent, and the builder leaves it empty.

What the filters can and cannot do

Under the hood a workflow is a saved search over the same index as the hiring signals API, so the vocabulary is the same:

• Regions become ISO country lists (“DACH”, “Nordics” and “North America” all expand), US states and Canadian provinces become region codes, cities become locations.
• Roles become job_title_or; technologies and domains become skills_or slugs. There is no free-text description filter in workflows, so the skill slug is the technology handle.
• Seniority uses five levels, and the builder includes unlabeled postings for non-executive levels, because most ads carry no seniority label at all.
• “Direct employers only” sets employer_type_not to agencies and brokers. Company-size words map to employee-count bounds. Named accounts go to company_name_partial_match_or.
• Employment type is applied with unlabeled postings included, since roughly a quarter of postings carry no type. “Remote” is a flag.

There are no salary or revenue filters. If you ask for one, the builder says so and configures the closest honest alternative instead of pretending.

Twenty-one templates when you would rather not type

Templates cover six jobs to be done. Sales and lead gen: companies adopting a technology, target-account hiring watch, ICP hiring surge, new GTM team build-outs, function build-out watch, executive and leadership hires, ATS-customer hiring signal, startup hiring watch. Recruiting: sourcing feed, agency BD leads. Competitive intel: competitor posting tracker, regional expansion watch, hiring-freeze warning, closed-role watch. Job boards and feeds: inventory webhook. Job seekers: personal, remote-only and contract alerts. Research: skill-demand digest, market trend report, salary benchmark alert.

Delivery: email, Slack or a signed webhook

Matches land in Signals, the same place saved searches from the original Signals launch live, and they reach you through one of three destinations:

• Email, either as each match is found or as a daily digest.
• Slack, into the channel you pick.
• Webhook, an HMAC-signed alert.matched payload with the same signing scheme as our job webhooks, so your pipeline can verify it.

“Instant” means within hours of a posting entering the index, not minutes, and the builder says so rather than overpromising. Matches are keyed on the first time a posting entered the corpus, so a reposted or backfilled ad does not fire twice. Evaluating a workflow costs no job credits; you pay only for postings you pull through the API.

Bring your own tools

The builder is itself an agent, and it accepts tools. Under Tools you can add any public MCP server, and from then on the builder calls it when it helps: pull the account list from your CRM before configuring a target-account watch, or check a company’s stack before settling on a skill filter. Every chat is saved under History, so a watch you refined last week is one click away. If you build agents yourself, the same index is available to them through our MCP server.

Plans

Every plan includes Workflows. Free accounts can run three, delivered by email. Builder accounts run ten and Scale accounts fifty, with Slack and webhook destinations on the paid plans. If you are on the API already, nothing changes: a workflow is a saved search that runs for you, and the postings it finds are the same records the API returns.

Open Workflows and type the thing you keep checking by hand. The builder does the rest.