Search & research

Tavily Web Search + Burrak AI

Give an agent the live web, with the links to check it by.

Connects with OAuth — revoke any time

A model knows what it read during training, cannot tell you when that was, and cannot cite it. Tavily closes both gaps: the agent searches, opens the pages and pulls their actual content, so what comes back is current and every claim can carry a URL you can click. It also changes what the other connectors are worth — research is the half of most jobs that has to happen before anything gets written into Notion, a doc or a CRM.

In the connector catalogue: Search the web, extract pages, crawl sites.

Things you could ask for

Written the way you would actually say them. An agent plans the steps itself.

  • Research these three competitors and write the comparison into Notion, with a source link on every claim.
  • Read this company’s pricing page and tell me what changed since the version I have pasted below.
  • Find what has been published about this CVE in the last week and separate what is confirmed from what is being repeated.
  • Crawl our documentation site and list every page that still mentions the old product name.
  • Every Monday, check what our four named competitors shipped last week and send me the list with links.

Search returns the open web, and a confident content farm sits in the results next to a primary source — the agent reads both. Ask for the link beside each claim rather than a tidy summary, and open the ones a decision rests on. The citation is not a formality here; it is the only part of the answer you can verify.

Connecting it

  1. 1Create a Burrak account — the free plan is enough to try this.
  2. 2Open Connectors and choose Tavily Web Search.
  3. 3Authorise Burrak in Tavily Web Search's own window. Your password never reaches us; we receive a token you can revoke whenever you like.
  4. 4Give an agent a task. It will use the connector when the work calls for it.

Tavily Web Search and Burrak, in practice

Is this different from the agent just knowing things?
Substantially. Model knowledge has a cutoff it cannot see past and no way to show its working, so it is at its worst precisely where you need it most — anything that changed recently, and anything you intend to act on. A searched answer arrives with the pages it came from, which makes it checkable rather than merely fluent.
Can it crawl any site I point it at?
Technically it will try, and that is a reason to be deliberate rather than a licence. Crawling is the right tool for sites you own or are responsible for — your docs, your marketing pages, a site you have permission to audit. For everyone else’s, search and read the pages you need; pulling a whole site because it was easy is how you end up rate-limited and, depending on the site, in breach of its terms.
Does searching cost extra?
You authorise your own Tavily account, so searches count against whatever plan that account is on — the connector does not bundle a search quota. Worth knowing before you set an agent a standing research task that runs every morning, since that is a recurring cost on Tavily’s side rather than a one-off.

Pairs well with

Agents get more useful the more of your workflow they can see at once.

No credit card required

Your AI assistant is waiting.
Put it to work today.

Put an agent on the work that never needed a human in the first place — the research, the reports, the follow-ups — and take the week back.

Now on Google Play — iOS coming soon

Free plan · No credit card · Cancel anytime · Works on Mac, Windows, Android & Web