Expand ATS coverage: Recruitee, Personio, Workable, Teamtailor (research-verified)

[prPMDev]

Context

Deep research pass on 2026-04-24 verified which job-data APIs are truly public (no auth, stable, permissive). Four clear wins surfaced — all simpler than BambooHR/Workday and together covering significantly broader geographic and SMB ground.

Full research in `notes/expansion-research.md` (gitignored).

The four adapters to add

1. Recruitee (#1 priority — cleanest pattern match)

```
GET https://{company}.recruitee.com/api/offers
```
Returns JSON. No auth. Mirrors Greenhouse/Lever/Ashby adapter pattern exactly.

EU SMB coverage, lots of remote-first startups.

2. Personio (#2 priority — EU coverage gap)

```
GET https://{company}.jobs.personio.de/xml?language=en
```
XML format (need to add XML parsing helper — one-time cost). No auth. Strong German/European coverage.

3. Workable — widget endpoint (#3 priority — breadth play)

```
GET https://apply.workable.com/api/v1/widget/accounts/{slug}
```
Returns JSON. No auth. Workable has historically had ~50K customers globally — huge SMB reach.

Note: this is the public widget endpoint, not the full Workable API (which requires OAuth). Fields are lightweight but sufficient for our schema.

4. Teamtailor (#4 priority — Nordic + EU strengthening)

```
GET https://{tenant}.teamtailor.com/jobs.rss?per_page=N
```
RSS 2.0. No auth. Scandinavian and European coverage.

Implementation pattern

Each adapter is one file in `src/adapters/`, follows existing Greenhouse/Lever/Ashby structure:

• `fetchX(slug)` — hit endpoint, normalize to unified schema
• `hasX(slug)` — HEAD request for detect_ats probing
• Register in `src/adapters/index.js`

Add registry file per platform in `registry/`. Seed with 5-10 verified companies each.

What this does NOT change

• Existing three adapters (Greenhouse/Lever/Ashby) untouched
• MCP server surface unchanged (same three tools, same envelope)
• No breaking changes to library API
• No npm publish needed until all four ship (single 0.3.0 bump)

Suggested sequence

Single focused day:

1. Recruitee (morning — direct pattern match, ~90 min)
2. Workable (afternoon — similar shape, ~90 min)
3. Teamtailor (requires RSS parsing helper — ~2 hrs)
4. Personio (requires XML parsing — ~2 hrs)
5. Test + verify live against 3-5 companies per adapter
6. Update README roadmap + docs/filters.md if any new gotchas
7. Bump version to 0.3.0, publish both packages

Related issues

• BambooHR adapter #2 (BambooHR) — not viable per research, worth closing
• Workday adapter #3 (Workday) — brittle, deprioritize in favor of these
• Duplicate job IDs for multi-office postings (same role, different locations) #17 (duplicate IDs) — will apply to new adapters too, watch during implementation

Acceptance criteria

• Each adapter has a test fixture + passing tests
• Registry seeded with 5-10 verified companies per platform
• README "Platforms" table updated
• MCP tool descriptions unchanged (adapters work through existing tool surface)
• Smoke test: real Claude Desktop query hits all 7 adapters successfully