jobatlas
Website · Oink personal job alerts · Repository · Job Atlas on Apify · Public Actor directory · Job integration catalogue · Simple and normalized tiers · Release evidence
Open schemas, integration recipes, and agent skills for Job Atlas job-search
Actors. The Actor implementations may be hosted services; this repository
publishes their interoperability layer, not a claim that every scraper is open
source.
Product guides, runnable examples, versioned contracts, and integration paths
are published at jobatlas.dev. The operating rules
for search discovery, measurement, and evidence-led content are documented in
the SEO program.
Start one of the four flagship Actors with the versioned first-run pack index; each pack labels its sample evidence and keeps exact-run and destination proof separate.
LinkedIn Jobs Scraper (No Cookies) + AI Enrichment
Get LinkedIn jobs ready for alerts, job boards, and spreadsheets. Keep full
descriptions and original links, skip jobs already returned by earlier runs,
and add AI enrichment or English translation when needed.
Start with five jobs and a small cost cap. Enrichment and
translation are optional paid extras.
The LinkedIn, Y Combinator, and EURAXESS Actors turn public job postings into
the same stable, source-linked record shape. The AI Job Search & Fit Scorer adds
a separate candidate-evaluation layer: it searches 10 developer-job sources or
accepts normalized jobs, then returns a ranked shortlist by default or a
complete drop/hold/failure audit when requested, without requiring a customer
model key.
Unofficial integrations. This project and its Actors are independently
developed. They are not affiliated with or endorsed by LinkedIn, EURAXESS,
the European Commission, or source-site operators. Public page access is not
authorization to crawl or republish personal data; review source terms,
licensing, privacy obligations, and applicable law before use.
Promoted Actor catalog
The job integration catalogue
accounts for the current Job Atlas copies, their live legacy counterparts,
explicit out-of-scope candidates, contracts, and client support. See the
client migration guide before changing an existing
Task, workflow, webhook, schedule, or destination.
| Actor | Best for | Key advantages | Availability and verification boundary |
|---|
LinkedIn Jobs Scraper (No Cookies) + AI Enrichment | Public LinkedIn job search | Find fresh jobs, suppress already-delivered matches, and send clean records with complete descriptions when available to alerts, trackers, job boards, or agents. Optional enrichment and translation stay off until selected. | Public Store Actor; latest is a mutable selector. API, MCP, n8n, Make, Python, and Agent Skill assets are maintained. Airtable is a destination projection only; Zapier is not claimed. See the current default build API. |
EURAXESS Jobs Scraper (No Login) + Full Descriptions | PhD, postdoc, fellowship, research, and faculty vacancies | Research domains, requirements, funding, deadlines, contacts, multilingual keyword expansion, strict filters, deduplication, optional enrichment, and translation | Public Store Actor; maintained integration selector latest, which is mutable. API, MCP, n8n, Make, Python, and Agent Skill assets are maintained. Airtable is a destination projection only; Zapier is not claimed; check Apify for the current default build. |
AI Job Search & Fit Scorer — 10 Sources + AI Matching | Candidate-specific developer-job shortlists | Search 10 public developer-job sources or score your own job list against a résumé or profile. Get a ranked shortlist with a 0–100 fit score, a 0–5 delivery score after hard-requirement checks, evidence, skill gaps, and links to the postings. No model key needed. | Public Store Actor; latest is a mutable selector. API, MCP, n8n, Make, Zapier, Python, and Agent Skill assets are maintained. No Airtable asset is claimed for the scorer. See current release evidence; $0.02 per returned shortlist row or retained non-failure audit row. |
Implementation guides: LinkedIn jobs ·
Y Combinator jobs ·
EURAXESS jobs ·
AI job-fit scoring ·
n8n workflows ·
Make blueprints
The three source-specialized scrapers target the six-root nomad-agent-job-v1
envelope, but source-specific
inputs, evidence, custom fields, deployment state, and
pricing must not be assumed interchangeable. See the
EURAXESS contract guide and the
integration compatibility matrix for the
exact boundaries.
The scorer consumes that canonical job envelope and returns the distinct
nomad-ai-job-fit-v1 evaluation contract. See the
AI Job Search & Fit Scorer guide for bounded API,
MCP, n8n, Make, and Zapier starters plus live-proof boundaries.
Public enrichment-quality benchmark
The open benchmark contract and scorer
measure exact final-record enrichment, unsupported fills, deterministic-field
preservation, provenance integrity, repeated-run completion, and selected-field
translation for LinkedIn and EURAXESS separately. The first human-verified
dataset is still being prepared, so the repository does not yet publish an
official accuracy percentage. The current evidence and its limitations are
documented in the benchmark status report.
The human review plan
defines qualified reviewer recruitment, blind double-labeling, adjudication,
service options, and the bounded first GT pilot.
Start with MCP
Connect only the tools needed by the current task. Both profiles use generic
call-actor so the exact build and cost caps are explicit:
https://mcp.apify.com?tools=fetch-actor-details,call-actor,get-actor-run,get-dataset-items,get-key-value-store-record
Inspect Actor details, then use generic call-actor with build latest for
LinkedIn, EURAXESS, Y Combinator, or the AI Job Search & Fit Scorer. The selector
is mutable, so confirm terminal success, verify the exact resolved build through
the Apify run API, validate the Actor-specific RUN-SUMMARY, and reconcile the
default dataset. The YC API and MCP examples are documented-only until
standalone assets are maintained and a dated live-readiness run proves them.
Or let the Apify CLI configure a supported client:
apify login
apify mcp install codex --tools fetch-actor-details,call-actor,get-actor-run,get-dataset-items,get-key-value-store-record
Replace codex with claude-code or cursor as needed. Then try:
Search for up to 20 remote or hybrid TypeScript developer jobs in Spain,
posted within the last 7 days. Do not translate or use AI enrichment. Return a
compact table and keep the canonical normalized records available.
See the complete MCP setup for ChatGPT, Claude, Cursor, and
Codex instructions and per-Actor compatibility gates.
Integration packs
| Priority | Pack | Workflow | Status |
|---|
| 1 | n8n | Daily alerts, normalized-job trackers, or a scored-shortlist Sheet | LinkedIn selector latest; EURAXESS and fit scorer selector latest |
| 2 | Make | Completed Actor run -> validated Google Sheets upsert | Task-owned complete inputs; exact scraper and fit-scorer builds required |
| 3 | Airtable | Import 32 fields and upsert on stable jobKey | Shared flat destination preset with linkedin and euraxess source choices |
| 4 | MCP | ChatGPT, Claude, Cursor, Codex, or another MCP client | Exact-build generic calls and Actor-specific output contracts for the maintained MCP examples |
| 5 | API/webhook | Custom job board, database, or internal application | Exact-build REST recipes; the scorer client is settlement-aware and validates legacy v3 plus current v4 result-policy and billing receipts |
| 6 | Zapier | Scheduled scored-job Sheet | Editor specification using scorer latest; not a published Zap or destination proof |
| 7 | Python parsers | Validate canonical JSON or create a destination projection | Source-specific validators plus the fit-row adapter |
The REST, MCP, n8n, Make, and Zapier directories also contain a distinct
AI Job Search & Fit Scorer pack using latest. Its destination uses
matchKey, not the source-only jobKey; see the
fit-scoring integration guide.
The fit starter explicitly sets resultMode: "shortlist" and
minDeliveryScore: 2. Use audit only when the destination genuinely needs
the complete drop, hold, and failure trail; those retained non-failure
decisions are billable.
Output policy
The Actor's canonical output remains the nested six-root
nomad-agent-job-v1 record:
schemaVersion, identity, data, custom, llm, raw
Integrations use a derived nomad-agent-flat-job-v1 projection. The flat view
is convenient for tables and no-code tools, but it does not replace the
canonical record. In particular:
- deduplicate on
jobKey (source:externalId), not job title;
- preserve the original record when downstream logic needs full requirements,
contacts, provenance, or the distinction between
null and [];
- never infer missing values from prose unless the Actor's opt-in enrichment
explicitly filled them;
- treat
raw: null as valid when includeRaw is false.
Fit-scoring integrations keep nomad-ai-job-fit-v1 as their canonical output
and derive a separate nomad-ai-job-fit-destination-v1 table projection. Do
not feed that projection into consumers expecting the six-root job envelope.
Agent skills
Repository-scoped Codex skills live under .agents/skills for the normalized
Actors. Claude Code users can install them into .claude/skills:
Set up the LinkedIn skill and the five scoped Apify MCP tools together in the
target project:
python3 scripts/setup_linkedin_monitor.py --client codex --target /path/to/your/project
Use --client claude or --client both when needed. Repeating the same command
keeps identical skill and MCP entries unchanged. The setup uses the dedicated
names apify_linkedin_jobs for Codex and apify-linkedin-jobs for Claude; it
preserves unrelated entries and fails closed if either dedicated name already
has different settings. It never embeds a token. Complete OAuth after setup
only after opening and trusting the target project in Codex; then run codex mcp login apify_linkedin_jobs. For Claude Code, use its /mcp menu. Restart an
already-open client after trust and authentication. A locally modified skill is
never overwritten unless --force-skill is explicit.
For skill-only installation:
python3 scripts/install_skill.py --skill linkedin-enrich-translate-normalize-scraper \
--client claude --target /path/to/your/project
python3 scripts/install_skill.py --skill euraxess-enrich-translate-normalize-scraper \
--client claude --target /path/to/your/project
See Agent skill setup.
Repository layout
.agents/skills/ Codex-compatible Agent Skills
benchmarks/ Public quality contracts, scorers, fixtures, and results
catalogue/ Versioned logical-product, deployment, and client inventory
docs/ MCP, contracts, and integration guidance
integrations/ Importable integration assets and shared projections
scripts/ Repository setup utilities
tests/ Offline contract and mapper tests
Security and responsible use
- Prefer OAuth for the hosted Apify MCP server. Never commit an Apify token.
- Use the least-privilege MCP URL above instead of exposing every Actor/tool.
- Review the selected source's terms, licensing, privacy requirements, and
applicable law for your use case.
- Do not publish personal data or raw descriptions to destination systems
without confirming the destination's retention and access controls.
Project status
LinkedIn 0.6 and EURAXESS 1.0 are public Store Actors. Y Combinator 1.0
and AI Job Search & Fit Scorer 0.1 are also public Job Atlas Store Actors.
Select latest in every maintained integration, retain the returned immutable
build identity, and verify destination behavior independently before enabling
a recurring workflow. Existing Actor-run evidence does not prove hosted/no-code
channel behavior or named-destination delivery.
License
The schemas, scripts, Agent Skills, examples, and documentation contained in
this repository are available under the MIT License, including for
commercial use.
The license applies only to files published in this repository. Hosted Apify
Actor implementations are not included here and are not licensed by this
repository.
Y Combinator startup jobs
| Actor | Best for | Contract and release |
|---|
| Y Combinator Jobs Scraper | Startup recruiting pipelines and recurring alerts | Six-root normalized records; exact release latest |
YC guide · Website · Agent Skill · Public YC v2 schema.
This normalized profile is separate from the legacy flat-export YC Actor.
It starts at $0.30 per 1,000 delivered jobs in Actor result events, set with
the October 2026 release. Destination templates for other sources are not YC-validated.
Quick setup for any public Actor
Install the generic Agent Skill to configure an Actor from its current schema and inspect its output. The setup guide explains bounded runs and result inspection. Use the job input examples and simple job recipes for schema-checked inputs.
SPAC filings on SEC EDGAR
SPAC Redemptions Scraper follows SPAC IPO, merger, vote and shareholder-redemption filings. The full guide includes date filters, output examples and a setup skill. Financial values stay unavailable when the filing does not establish them.
