TinyFish is launching TinyResearch and TinySet for founders, operators, and the agents they already run. TinyResearch takes one hard question, investigates it across public pages, and returns a cited report that opens with a summary and comparison tables so the answer can go into a meeting. TinySet takes a plain-English description of the same facts across many companies, places, or products, lets you edit the columns it proposes, and builds a dataset where each value comes from a page an agent actually read. The Palo Alto company already offers search, fetch, cloud browser sessions, and web agents behind one API, with Search and Fetch free. It productized these two after its own team kept building the same reports and datasets by hand for customers, stitching each one from those tools. General deep research often returns a long writeup that still leaves the decision open, or a table that covers the first few dozen rows and drops the rest. TinyResearch keeps another round going until the gaps close. TinySet splits a large ask into smaller pieces so later rows are researched the same way as the first. A Deep report usually takes 5 to 20 minutes, and you can run up to five in parallel. Both sit on the same account and wallet as the rest of the platform, and they are available in the playground, as a ChatGPT plugin and a Claude connector, and through API, MCP, CLI, and SDKs. A Deep TinyResearch report costs $2 and a Max report costs $3, charged only if it finishes. A TinySet row costs 10 cents. New accounts start with $8. Both work on the public web only for now. A dataset is a snapshot you rerun for fresher data, and a run collects up to 250 rows.