Exa integration

Exa AI search & web-retrieval automation

25 operationsAI

Exa is a search engine built for AI, with neural search, direct answers, and structured research over the live web. The NoClick Exa node calls that API directly, so a workflow can answer a question with cited sources, find pages similar to a URL, pull the full contents of any page, and build Websets that collect and enrich results on a topic. It also runs Exa's research agents and sets up monitors that watch the web and re-fire on new results. Add it to the canvas to turn open-web research into a step your automation runs on its own.

What you can automate with Exa

Exa operations

Triggers1

Search4

Agent4

Monitors7

Websets9

Templates using Exa

Exa Web Research Monitor

Turns each Exa monitor run into one Slack digest of the few results worth reading, each with a line on what it changes for you, and the rest dropped with a reason. Exa monitors are good at finding pages that never contain the words you searched for, which is the reason to run them and the reason a raw result feed is unreadable. This agent takes each run, throws out what is off brief or already covered, opens only the results that earn the cost of opening, and posts a single Slack digest saying why each survivor matters. What gets watched stays configured in Exa; what reaches your team is the judgment this adds on top. ## What people use it for - **Research without the reading** - Somebody on your team is already skimming forty links a week to stay current on a market. This does the skimming and hands over the four that changed something. - **Nothing arrives twice** - Earlier runs are checked before anything is written, so a story that travels for three days across a dozen outlets reaches you once, from whichever source got there first. - **Papers your keywords would miss** - Academic and engineering write ups routinely describe your problem using none of your vocabulary. Semantic monitoring catches them, and this layer decides which are worth a working day. - **The dropped count keeps you honest** - Every digest ends with what was discarded and why. Read a fortnight of those lines and you know exactly where the brief is too loose, which no unexplained silence would ever tell you. ## Before you fork **What has to exist in Exa before we fork this?** An Exa API key, at least one monitor already created there with its query and schedule, plus whichever Slack channel should carry the digests. The agent reads monitors and never creates them, so shaping the watch list stays a deliberate act you perform in Exa. Test Run stages a batch of results against your brief so you can read a digest before any of it is live. **What drives the cost of a run?** One agent run per batch of results, plus the Exa content calls, and content is the metered part. Because only the shortlist is opened, a tight brief is also the cheapest brief. A monitor returning forty items a night costs noticeably more than one returning five, so narrow the query in Exa before you widen the brief here. **How do we keep it from drifting off topic?** The brief is the whole control surface, and nothing else influences what is kept. If two weeks of digests show it letting adjacent material through, name that material in the brief as noise and it disappears from the next run. If it is dropping things you wanted, the dropped counts tell you which category to loosen.

94004 nodes

Questions

Yes. The Answer operation sends a question to Exa and returns a written answer along with the sources it drew from. If you need the raw material instead, Get Contents pulls the full text of any page so a later step can process it.

The Exa node authenticates with an API key. Create a key in your Exa dashboard under API keys, paste it into the node's credentials, and every operation in the workflow uses it. Connecting the credential also registers the monitor webhook automatically, so the trigger works without extra setup.

Yes. Create a Monitor for a search you care about, and the On Monitor Results trigger fires the workflow each time that monitor produces new results. From there you can summarize the results, store them, or route them wherever you need.

It creates a Webset from a search, adds enrichments to fill in extra fields on each item, and lists the items once Exa has gathered them. You can also create a Webset webhook so later steps run as the collection updates, and list a Webset's events to track what changed.

Yes. You can create an agent run to start a research task, check its progress with get agent run, list past runs, and cancel a run that is still going. That lets a workflow hand off longer research jobs to Exa and pick the results back up when they land.