Site Search 360
Web PlatformConnect AI agents to Site Search 360, a hosted site search service for websites. Agents run searches, fetch autocomplete suggestions, index or remove content, and read index status and query analytics so on-site search stays fresh and measurable.
What This Integration Enables
Site search is the one place on a website where visitors tell you, unprompted and in their own words, what they came for. Most teams treat that box as a widget they installed once. Site Search 360 is worth connecting because it treats it as an index you own, and this connector gives agents both halves of that: the ability to keep the index honest, and the ability to query it.
Two of those halves run against different endpoints with different credentials, which is unusual enough to plan around. Search and Get Suggestions authenticate with the Site ID and are the public-facing reads, capped at 100 results per request with a default of 10. Everything that changes or inspects the index, meaning Index URL, Index Content, Index Content Batch, Delete Document, Delete Documents By Pattern, Get Index Status, List Indexed Content, and Get Query Analytics, authenticates with the API key against the management endpoint. The practical consequence is that a flow can safely run search queries in a lot of places while the indexing actions stay in a smaller number of controlled ones.
The indexing side offers a real choice rather than one path. Index URL asks the crawler to go and fetch a page, which is right when the content is public and the crawl settings already do what you want. Index Content and Index Content Batch push a document directly with its title, body, content group, data points, filters, keywords, and ranking boost, which is right when the content sits behind a login, lives in a system the crawler cannot see, or needs metadata the page does not carry. Batch handles up to 100 documents per request. Search itself is more than a visitor-facing feature once an agent can call it: it is a retrieval tool over the site's own published content, which makes it useful anywhere a flow needs to find the page that answers a question.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Publishing and indexing on the same run
An article goes live in WordPress. Rather than waiting for the next crawl, the agent calls Index URL for the new page the moment it publishes, then calls Get Index Status to confirm the count moved. For the parts of the site the crawler cannot reach, meaning a knowledge base behind a login or a product catalog rendered from a database, the same run pushes documents directly with Index Content Batch, supplying the content group and the data points the search results template expects. Search results reflect the site as it is now rather than as it was at the last crawl, and nobody had to remember to press a button in a control panel.
The site's own content as a retrieval tool
A support agent handling a ticket in Zendesk needs the page that actually answers the customer's question. Before drafting anything, the flow calls Search with the customer's own words, reads the ranked results and their snippets grouped by content group, and pulls the two or three pages that match. The reply is built from published content with real URLs attached rather than from a summary that might be a year out of date. Get Suggestions serves the same purpose in a live context, returning autocomplete suggestions and grouped previews as a partial query is typed, which is how a flow narrows a vague request before committing to an answer.
An index that gets audited instead of accumulating
Monthly, the agent walks the index with List Indexed Content, filtering by HTTP status to find documents whose URLs no longer resolve and by content type to find things that were never meant to be searchable. It pairs that with Get Query Analytics, which reports the timeline of search volume over a period defined by start and end timestamps, so the team can see whether searching went up or down around a content change. Note what that analytics action does and does not do: it reports how much searching happened, not which individual query failed, so the honest use is trend detection rather than a list of content gaps. The output is a report naming the sections with dead entries and the volume trend around each recent publish, and it goes to the content owner with a proposed cleanup.
Human-in-Loop Highlight
The action that decides where the gate goes is Delete Documents By Pattern. It removes every indexed document whose URL matches a supplied regular expression, and it is the only efficient way to clear a retired section of a site. It is also a single field where a small mistake has no floor. A pattern meant to match /blog/2019/ that is missing an anchor matches every URL containing those characters anywhere. A pattern intended as a literal path where a dot is not escaped matches considerably more than the author expected. The removal is permanent, and the recovery path is not a call in this connector: documents come back when the crawler fetches them again, which for a large site is hours of degraded search results while every visitor who uses the box gets nothing.
So the agent does the work that makes the decision safe and hands over the one field that carries the risk. It calls Get Index Status for the current document count, calls List Indexed Content filtered to the URL substring in question, and posts to Slack: "Proposed index cleanup for the retired 2019 blog archive. Pattern: ^https://example\\.com/blog/2019/. Matches 284 of 9,140 indexed documents. First 15 matched URLs listed below, plus the 3 closest URLs that do not match so you can see where the boundary falls. This removal is permanent and matched pages return only on a re-crawl. Approve this pattern, or send back an edited one?" The person can approve it, edit the regex, or reject it, and the agent executes whichever came back.
That framing is the difference between an approval message and a real gate. The agent is not asking permission to do a chore. It is presenting the exact blast radius of a pattern, in documents and in URLs, at the moment before the pattern is applied, which is the only moment the information is useful. Delete Document, by contrast, takes one exact URL and removes one document, which is precise enough to belong in an automatic path when a CMS confirms a page was deleted. The gate belongs on the wide lever, not the narrow one, and this is human-in-the-loop as Orchestration as a Service means it: the agent does the counting and the matching, and the person owns the irreversible call.
Agent Capabilities
10 actionsSearch
2- Search Runs a full text search against the configured site index and returns ranked results grouped by content group, with pagination, sorting by a data point, filtering, content snippets, and query term highlighting. The maximum limit is 100 results per request and the default is 10. Beyond powering the site's own search box, this is a retrieval tool over your published content that any flow can call.
- Get Suggestions Returns autocomplete query suggestions and grouped result previews for a partial query, suitable for a search-as-you-type box. Results can be grouped by content group and the number of suggestions capped, which makes it useful for narrowing a vague request before a flow commits to an answer.
Indexing
5- Index URL Pushes a single URL to the crawler so the page is fetched and added to or refreshed in the index, processed according to your control panel crawl settings. The right call the moment a public page publishes or changes, rather than waiting for the next scheduled crawl.
- Index Content Indexes a single document from supplied JSON content instead of crawling, with full control over title, body, content group, data points, filters, keywords, and ranking boost. Used when the content sits behind a login, lives somewhere the crawler cannot see, or needs metadata the rendered page does not carry.
- Index Content Batch Indexes up to 100 documents in one request, each using the same structure as Index Content. The efficient path for bulk imports, catalog syncs, and any first load of a system the crawler was never going to reach.
- Delete Document Removes a single indexed document identified by its exact URL. Precise enough to run automatically when a CMS confirms a page was deleted, because the blast radius is exactly one document.
- Delete Documents By Pattern Removes every indexed document whose URL matches a supplied regular expression. The efficient way to clear a retired section and the widest lever on the connector, because the removal is permanent and a slightly wrong pattern takes far more than it was aimed at.
Site Management
3- Get Index Status Returns the current status of the index, including how many pages are indexed. Used to confirm that a run of indexing calls actually landed, and as the denominator when a person is being shown how much of the index a proposed deletion would take.
- List Indexed Content Lists the documents currently in the index with their URL, content type, and crawl status, filterable by URL substring, content type, or HTTP status, and paginated from 1 to 100 per page. The read behind index audits, stale-entry detection, and the preview attached to any bulk removal request.
- Get Query Analytics Returns a timeline of search query volume for the site over a period defined by start and end UNIX timestamps in milliseconds. This reports how much searching happened rather than which individual queries were run, so treat it as trend evidence around content changes rather than a list of failed searches.
Frequently Asked Questions
What can FlowRunner do with Site Search 360?
FlowRunner agents can run Search, Get Suggestions, and Index URL in Site Search 360, plus 7 more actions.
Does connecting Site Search 360 to FlowRunner require OAuth?
No. Site Search 360 connects to FlowRunner with an API key, no OAuth flow required.
Can Site Search 360 trigger a FlowRunner workflow automatically?
Site Search 360 doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Site Search 360
$100 in credits. No card required. Connect in minutes.