REST API · Automation integrator

Automation integrator → machine builders in Germany: what one real Prosp run returned

By the Prosp team · run on 2026-10-03 in production

Run date

2026-10-03

Job duration

4 s

Results came from

Prosp knowledge base

Delivered

8 of 8 requested

Rejected candidates

19

Discarded after target

83

The scenario

An automation integrator wants mid-sized German machine builders, including special-purpose machinery makers, that could hire it to automate their machines. The buyer is a profile we set up to test this market, not a customer. The run is real, and every company on this page came back from it.

Buyer

An engineering firm that integrates control and automation systems into industrial machinery. An illustrative profile, not a named customer.

Target

Medium and large German machine builders and special-purpose machinery manufacturers (Maschinenbau, Sondermaschinenbau).

The exact request

Sent to the REST API as POST /api/v1/discoveries:

curl -X POST 'https://api.prosp.cc/api/v1/discoveries' \
  -H "Authorization: Bearer $TOKEN" \
  -H 'Content-Type: application/json' \
  -H "X-Idempotency-Key: $(uuidgen)" \
  -d '{
  "query": "Find mid-sized German machine builders and special-purpose machinery manufacturers that could use an automation integrator",
  "industries": [
    "manufacturing"
  ],
  "locations": [
    "Germany"
  ],
  "company_sizes": [
    "medium",
    "large"
  ],
  "keywords": [
    "Maschinenbau",
    "Sondermaschinenbau",
    "machine builder"
  ],
  "maxResults": 8
}'

How this run was actually started: through Prosp's internal job endpoint, with exactly these criteria and a target of 8 companies, under an internal exploration account whose results never reach a CRM. That path does not charge credits. Through the public API the same body is sent as POST /api/v1/discoveries, which debits 100 credits; you then poll GET /api/v1/discoveries/{jobId} and read GET /api/v1/discoveries/{jobId}/prospects.

What came back

Checking the registry ids

Why candidates were rejected

Not counted above: 83 more candidates were still being processed when the target was reached. They were discarded without being evaluated (terminal_job_discarded), so they say nothing about quality either way.

What went wrong

• Both register ids on this run point at the wrong company, according to each company's own Impressum. pfmgermany.de was matched to a Saarbrücken entry for a company based in Lower Saxony, and pfaff-industrial.com to a namesake registered in Ulm. A buyer who trusted those ids would pull the wrong company's filings. We are fixing the matching that let them through.
• This was not a fresh search. All 8 companies came from Prosp's knowledge base, which is why the job took 4 seconds (3.8 s measured). Do not read that number as search speed.
• Fit: Automatisierungstechnik Berger and Broetje Automation describe themselves as automation providers, so for an automation integrator they look more like competitors than customers. MKW offers both special-purpose machines and automation.
• 4 of 8 rows have no company size (MKW, Konrad Seidler, Broetje Automation, PFM Germany), and 2 have no city (MKW, PFM Germany), although PFM's own description names Hollenstedt.
• 4 of the 8 had already been delivered to the same exploration account by earlier runs.

How to read this run

All eight companies came from Prosp's knowledge base, so the job ended within seconds and the live search it started in parallel never got to deliver: 83 of its candidates were still in flight and were discarded without being evaluated. Only 19 candidates were actually rejected.

The closest fits for an automation integrator are the companies that build machines and would buy automation for them: Boschert (CNC sheet-metal machinery), Konrad Seidler Maschinenbau (vertical turning lathes and machine retrofits), KLEIN Maschinenbau (tape-processing and special machines), PFM Germany (packaging machinery) and Pfaff-Industrial (industrial sewing, welding and joining systems). That is five of eight, and the two with the most to check before outreach are the two whose register ids are wrong.

What this market can and can't verify

• Registers: not with a simple lookup. The German Handelsregister has no free read-only API where you can paste a number and get a record; its search runs as a browser session. We did not query it for this article.
• Impressum: yes. German law requires a business website to name its legal entity and, if registered, its court and register number. On 2026-10-03 we opened the Impressum of both companies whose rows carried a register id. Neither matched: one names a company in Lower Saxony and gives no register number, the other gives a different court and a different number.
• So for Germany, check every delivered register id against the company's Impressum before you rely on it. On this run that check caught two out of two.

Cost and time

• The job took 4 s (3.8 s measured) because it was served from the knowledge base. Fresh runs in the same series took between 9 min 31 s and 14 min 13 s. At submission the API estimated a median of 565 s.
• A discovery costs 100 credits through the REST API, the MCP server or the Telegram bot. This run was started on an internal account and was not charged.

How to reproduce this run

• REST API: get a token, then send the body above to POST /api/v1/discoveries with an X-Idempotency-Key header. See the API reference.
• ChatGPT or Claude: connect the MCP server and call find_prospects with the same criteria.
• Telegram: run /find in the bot. See the Telegram guide.

Your run may be served from the knowledge base too, or may need a fresh search if the knowledge base holds fewer matches by the time you run it.