Gistly

AI agents need the live web, and the live web is slow, messy and expensive to read well. Every team building an agent was rebuilding the same search, fetch, extract and summarise pipeline under deadline, and none of them wanted to maintain it afterwards.

We built it once and put it behind one endpoint with a published price list.

A good answer needs the right pages, not the first pages, and some of those pages only exist after JavaScript runs. The question and the reader are often in different languages. An agent needs to know what each call cost before it makes the next one.

Each of those is a product decision as much as an engineering one, and each had to be made in a way a developer could read in the documentation in a minute.

Gistly has two tools: a search tool that discovers, fetches, extracts and ranks sources, then writes one answer with the sources listed, and a fetch tool that returns any page as clean Markdown. Around them sit a prepaid wallet with the charge in every response, a portal with keys, usage and a playground, and a contract-first API whose reference and client are generated from the code.

Depth is a dial with four settings: quick, standard, deep and max, each at a fixed price. The answer can be asked for in any of twelve languages while the search stays in the question's own.

• Search: one written answer with sources, four depths from quick to max
• Fetch: any page as clean Markdown, JavaScript rendered when needed
• Twelve answer languages; the search stays in the question's language
• The cost and your balance in every response, in the body and the headers
• A customer portal with keys, usage, wallet and a playground