Package Information
Available Nodes
Documentation generating…
Documentation
n8n-nodes-serpdive
An n8n community node for SERPdive, a web search API that
returns the extracted text of each page, not a list of links with two-line
descriptions.
That difference is the point in n8n. A search that gives you snippets needs a second
step to fetch each page before anything downstream can read it. This one answers in a
single call, which matters most when the node is wired to an AI Agent: the agent reads
the page in the same turn instead of planning a round trip.
Installation · Credentials ·
Operations · Using it as an agent tool
Installation
In n8n, go to Settings > Community nodes, select Install, and enter:
n8n-nodes-serpdive
For a manual install, see n8n's community nodes
documentation.
Credentials
Create a key at serpdive.com/dashboard/keys. The
free tier needs no card.
In n8n, add a SERPdive API credential and paste the key. Use Test to confirm it
works: the check runs one search on the free model, so it costs nothing even on a paid
plan.
Operations
Search is the only operation. It takes a query and returns ranked results.
| Field | Default | Notes |
|---|---|---|
| Query | — | Required, up to 300 characters |
| Model | mako |
krill, mako or moby, see below |
| Include Answer | off | Also return a written answer synthesised from the results |
| Max Results | 10 | Between 1 and 10 |
| Split Results Into Items | off | One item per result instead of one item for the whole response |
Each result carries url, title, content and, where the page states one, date.content is the extracted page text.
Split Results Into Items is off by default so that nothing is dropped without being
asked. Turn it on when the next node loops over results. Leave it off when you need the
answer or the response timing, which belong to the response as a whole.
Models
krill is the free tier. It is unlimited under fair use, with no monthly credit
allowance that runs out, and it allows one search at a time per account: a second
call made while one is in flight is refused rather than queued. For a workflow that
searches step by step, that is the natural shape. If you fan several searches out in
parallel, use a paid model.
mako extracts more deeply and allows parallel searches. moby reads the most pages.
Using it as an agent tool
The node declares usableAsTool, so it can be attached directly to an AI Agent
node's tool port. Give it a query expression and the agent decides when to search.
Because a result carries the page text rather than a snippet, the agent can cite what
it read without a follow-up HTTP Request node. The trade is token volume: a payload
with page text costs more to read than a list of links, and on a paid model withMax Results at 10 it can be substantial. Lower Max Results if the model's context
bill matters more than coverage.
Compatibility
Tested against n8n's current node API (n8nNodesApiVersion: 1). Requires an n8n
version that supports community nodes.