n8n-nodes-linkfinderai

An n8n community node for LinkFinder AI —
turn a name, email, company, or LinkedIn/Instagram URL into contact and company data,
directly inside a workflow. No more wiring up the HTTP Request node by hand.

This wraps LinkFinder AI's single-endpoint API (POST https://api.linkfinderai.com,
one type field per operation — see the API docs)
into proper n8n Resource/Operation dropdowns, with built-in handling for the API's
sync/async split (most lookups resolve inline; linkedin_profile_to_linkedin_info
is always async and returns a job to poll).

What's included

• Credentials — LinkFinder AI API, just an API key (find it in your dashboard
  under Settings → API Key). Includes a live credential test.
• Node — LinkFinder AI, covering all 20 operations across 7 resources:
  • Lead: Full Name → LinkedIn URL, Full Name → Email, Email → LinkedIn URL
  • Company: Name → Website / Phone / Email / Employee Count / LinkedIn URL,
    Name → Employees, Domain → Employees
  • LinkedIn Profile: URL → Full Info (async), URL → Email, URL → Phone
  • LinkedIn Company: URL → Info, URL → Employee Count, URL → Employees
  • LinkedIn Post: URL → Reactions
  • Instagram: Profile URL → Info
  • Discovery: B2B Data Lookup
  • Job: Check Status — poll a job by Job ID or Poll URL

Each operation shows its real credit cost in the dropdown, and each has its own
labelled Input field describing exactly what to paste in.

Looking someone up by name

The two name operations — Lead → LinkedIn URL and Lead → Email — take four
fields rather than one: Full Name (required), Company, Location and
Job Title. They are joined into the single string the API takes, with blanks
dropped, which is exactly what the LinkFinder app itself sends.

Map as many as your data has. The extra three cost no additional credits and they
are the difference between one match and thousands: John Smith matches an enormous
number of people, John Smith Acme Berlin VP Sales matches one — and Lead → Email is
7 credits a row either way, charged whether it finds the person you meant or a
stranger with the same name.

A name coming out of a CRM as Doe, John is reordered to John Doe automatically. A
company with a comma in it (Gates, Foundation) is left alone.

The operation list is generated, not hand-written

nodes/LinkFinderAi/generated/operations.ts is built from
integrations/catalog/operations.json, which is
itself built from the root openapi.json. npm run build regenerates it.

This replaced a hand-maintained map that had drifted from the API: four operations
were missing, Instagram posted a type the API does not accept, and the employee
filters were wired to one of the three operations that take them. Do not edit
anything under generated/ — change the catalog and rebuild.

Credits are not uniform

URL → Phone costs 50 credits; URL → Full Profile Info and URL → Email cost
10; Full Name → Email costs 7; most company lookups cost 1. Employee
lists bill 0.5 credits per employee returned. Every call is charged, including one
that finds nothing — which matters in a workflow that runs unattended over a list.

How the async handling works

Every operation has a Wait for Completion option (on by default). When a call
comes back with a job_id instead of an immediate result, the node polls
internally — growing delay, capped wait window (Max Wait Time, default 25s;
60s for the five operations that are always async — linkedin_profile_to_linkedin_info,
the three employee-list operations, and the AI lead search — since 25s usually isn't
enough for those).

If the deadline passes before the job finishes, the node returns
{ processing: true, job_id, poll_url } instead of failing. Wire that into an
n8n Wait node followed by a LinkFinder AI → Job → Check Status node,
looping until processing is false — the same pattern documented for
Zapier/Make and used by
LinkFinder's own HubSpot integration.

Turning Wait for Completion off skips the internal poll entirely and always
returns the job info immediately, if you'd rather build the wait loop yourself
from the start.

Local development / testing

This hasn't been run against a live n8n instance from this environment — do
that before publishing. n8n's own guide for testing a community node locally:
https://docs.n8n.io/integrations/creating-nodes/test/run-node-locally/

cd n8n-nodes-linkfinderai
npm install
npm run build          # tsc + copies the icon into dist/
npm link                # registers this package globally

# in your n8n install / ~/.n8n:
npm link n8n-nodes-linkfinderai
n8n start

Then in the n8n UI: add a LinkFinder AI node, create the credential with a
real API key, and try a cheap operation first (e.g. Company → Name → Website)
before testing the async LinkedIn profile lookup.

Publishing

Once verified locally:

npm run build
npm publish   # package name must stay prefixed "n8n-nodes-" for n8n's
              # community-node installer/search to find it

After publishing, submit it to n8n's verified community nodes program
so it shows up as a first-class result in n8n's in-app node search rather than
requiring users to install it by exact package name.

Notes / things to double-check before publishing

• credentials/LinkFinderAiApi.credentials.ts's test request calls
  company_name_to_website with "Tesla" — that's a real 1-credit call against
  production, used only when a user clicks "Test" on the credential. Confirm
  that's the credit-cost tradeoff you want for a credential test.
• Error messages in assertOk() are deliberately generic per HTTP status
  (401/402/422/429/5xx) rather than always surfacing the API's message field —
  adjust if you'd rather always pass the raw API message through.
• No rate-limit-aware batching across items yet — a workflow item list of, say,
  200 rows will fire 200 sequential requests (with the built-in 429 backoff per
  request, but no concurrency control). Fine for typical workflow sizes; worth
  revisiting if this becomes a common complaint.