Every agent in Boldstream can now read any public web page directly during a chat. Drop a URL — a competitor's pricing page, a news article, a brand's About page, a piece of editorial coverage — and the agent pulls the live content into the conversation. No copy-paste, no "I can't access that link," no needing to summarise the page yourself first.
What this unlocks
Competitor checks on demand — paste a competitor's homepage and ask the agent how their positioning compares to yours, or what's new on their pricing page since last quarter.
Reading the news the agent is reasoning about — share an article URL and ask "is this a moment to react?" — the agent reads the actual coverage, not a guess of what it might say.
Pulling claims from your own pages — point at your own About page or a campaign landing page and ask the agent to extract the claims, the audience promises, or the proof points exactly as written.
Trade-press monitoring — drop a trade-publication URL into a PR or comms agent and have it triangulate the angle.
Validating a writing brief — ask a content agent to read a top-ranking page on the topic before drafting, so the output reflects current best-in-class rather than a stale snapshot.
How to use it
There's nothing to switch on — every agent has the capability by default. Just include the URL in your message:
"Read https://example.com/pricing and tell me how their tier structure compares to ours."
"Pull the claims from https://example.com/about and find which we could counter with our own evidence."
"Here's the article: https://example.com/news/article — what's the angle, and is there a response worth drafting?"
The agent fetches the page in real time, reads the content, and reasons about it the same way it reasons about any other context you provide.
What's still off-limits
Pages behind a login — the agent reads what a public visitor would see, so anything gated stays gated.
Sites that block automated readers — a handful of high-profile sites refuse all programmatic access; the agent will tell you when it can't read a specific URL rather than guessing.
PDFs and other non-HTML formats — for those, use the document upload flow inside the chat composer.
The shorthand: if it's on the open web and a human could read it, the agent can now read it too. Try it on your next competitive sweep — the difference between "summarise from memory" and "read the actual page" is the difference between guesswork and intelligence.