Templates / Enrichment
Web Agent
Web Agent fills in facts you would otherwise google by hand — a founder's name, a funding round, something written about the company elsewhere. It searches the web rather than one site.
What it isEnrichment
An enrichment column. It reads a column you already have and writes a new one beside it.
InputWhat you give it
A column that already holds the value this reads.
OutputWhat comes back
| Task | Result | Steps taken |
|---|---|---|
| Find the cheapest plan with SSO | Business — $89/user/mo | 4 |
| Check whether the trial needs a card | No card required | 3 |
| Get the support response-time SLA | Not published | 6 |
Illustrative. The browser-driving agent — it navigates, clicks and reads, for the jobs a single page cannot answer.
How it worksSix things you decide
- Ask what you would type into GooglePhrase it as the fact you want, and mention the company with a {Column} placeholder so each row searches for itself.
- Add fields for structured outputOptional — skip it for one free-form answer. Each field becomes its own column: name it, pick its type, and use Output description to tell the AI what belongs there.
- Pick a modelUnder LLM model settings. The list is whatever you've connected — an OpenAI, Anthropic or Google AI key from Settings, or Claude Code (local agent) with no key needed. A harder fact to track down does better on a stronger one.
- Turn on vision only if neededUnder Advanced settings. Helps when the fact is easier to see than to read, but adds tokens on every step. Text-only is much cheaper and usually enough.
- Give it roomUnder Advanced settings. Max steps is how many search-and-read turns it may take before it must answer.
- Pick how it searchesUnder Advanced settings. Browser search then scrape is the thorough route. Native web search is faster and needs no browser, when the session supports it.
FAQQuestions people actually ask
Why is Site Agent sometimes better?
When the answer is definitely on the company's own site, Site Agent goes straight there instead of paying for a search first.
What goes in Output description?
It's the instruction the AI gets for that one field — you never see it again, the model does. Say the prompt is "Find funding information about {Company}." One field: Column name "Last Round", Output description "The most recent funding round type — Seed, Series A, Series B." Another: Column name "Amount Raised", Output description "The dollar amount raised in that round, formatted like $5M — use the total if multiple investors are listed." Leave it blank and the AI just gets "extract value", which drifts in format from row to row.
Multiple fields, or separate agent columns?
Group fields that are likely to be found together, not just ones that are topically related. One run gets a fixed number of search-and-read turns no matter how many fields you ask for, so facts that turn up in the same search are nearly free to add on. Facts that need genuinely different research, like a funding round versus a competitor's name, share that same fixed budget, so the agent may settle for N/A on the harder one instead of chasing it properly. Split those into separate columns instead.
Does it use any of my logins?
No. Every page it reads is a stateless, signed-out fetch, and the search step runs in its own hidden browser session that only remembers cookie-consent choices — never an actual Google login. Your Reddit / Apollo / Facebook / LinkedIn connections from Settings are separate accounts scoped to those specific collectors — Web Agent never touches them.
What happens if Google blocks the search or asks to sign in?
It tries to recover, then fails cleanly rather than waiting on you. First it auto-dismisses a plain cookie-consent wall. For an actual CAPTCHA, it tries up to 3 fresh IPs if you have a rotating proxy set — a new IP is usually enough to avoid the challenge. If that doesn't clear it, the run never pops a sign-in window (that only happens on manual, non-agent searches) — this search just fails, and the agent falls back to whatever page it already had or answers N/A for that row. Without a proxy, a real CAPTCHA generally means that search fails outright, since consent-dismiss alone won't beat a bot challenge.
What does "Auto-run on new rows" do?
Runs this column by itself whenever new rows arrive — no clicking Run. It only picks up new arrivals; rows already in the table when you turn it on are left alone. It only fires while the app is open — closing it pauses everything until you reopen.
What does "Add run condition" do?
Skips rows where a condition on another column isn't met — e.g. only run if Email is empty. Skipped rows cost no actions, on both manual and auto runs.