Live web search

OpenPerplex runs a search query against current web results and returns ranked hits your application can act on. A Beam agent fires a query when a task needs fresh information, reads the returned results, and applies your rule for which sources or recency thresholds are acceptable before saving them to the record. Queries that come back with nothing usable, or hits from sources your policy distrusts, get routed to a person for a decision. The agent operates within your account limits and only where the connector and permissions allow, so search stays inside the boundaries you set.

Citation-backed answers

Beyond raw links, OpenPerplex can return an answer with the sources it drew from attached. A Beam agent asks a question tied to an open task, reads the answer and its citations, and checks them against an approved rule about which claims need a named source. Answers that pass get written into the ticket or brief, while answers that cite weak sources, or that touch a sensitive topic your policy reserves for people, are handed to a reviewer. This keeps automated research traceable and gives humans the final call on anything consequential or disputed.

URL content retrieval

OpenPerplex can fetch and read the contents of a specific page so downstream steps work from the real text, not a guess. A Beam agent supplies the URL, retrieves the extracted content, and applies your rule for length, language, or required sections before storing a summary or key fields. Pages that fail to load, sit behind a paywall, or return content unrelated to the request are flagged for a person instead of being used blindly. The agent stays within account and connector permissions, keeping retrieved text defensible and tied back to its source.