Frequently asked questions.

Keldura turns the sources you follow into living AI knowledge bases. It imports videos, websites, PDFs, documents, audio, and recurring channels, organizes them into Library collections, and lets you ask questions whose answers cite the source material. The same knowledge can be published as a bot on Telegram, embedded on your website, or exposed to AI agents.

Yes, there is a free tier. Every account gets a daily credit allowance that refills automatically, enough to try the full workflow. Paid plans raise the allowance and unlock extras like the website chat widget. Current plans and prices are on the pricing page.

YouTube videos, playlists, and channels; Instagram reels and profiles; websites; PDFs including scanned ones; Word documents; plain text and Markdown; images; audio and video files; Google Drive documents; and content forwarded through Telegram. Everything lands in one Library.

Keldura transcribes content in its original language and can translate transcripts and summaries into more than 90 languages. You can also chat with your knowledge base in your own language regardless of the source language.

No. Keldura uses existing captions when they are good, and otherwise transcribes the audio with speech-to-text models. Videos without captions work fine.

Answers cite the source material they are based on, with timestamps for video and audio. You can open any citation and check the original. If a collection does not contain the answer, a well-scoped bot says so instead of inventing one.

Yes. Your Library is private to your account unless you explicitly share a collection, a document, or a bot. Content is processed by AI providers to produce transcripts, summaries, and answers. See the privacy policy for details.

Yes. Collections can be shared with specific people or by link, documents and briefings have public read-only share links, and bots can be published to Telegram or embedded on your website. Every share can be revoked.

Yes. Any collection can be exported as a ZIP containing transcripts, summaries, and metadata. Individual transcripts and documents can be copied or downloaded, and the developer API offers programmatic access.

You can delete items and collections at any time; deletions offer a short undo window before becoming permanent. To delete your entire account and its data, contact support and we will take care of it.

A collection is a folder in your Library and the unit Keldura works with: you chat with a collection, share a collection, monitor sources into a collection, and give bots collections as knowledge. Group related sources together and Keldura answers across all of them.

Chat is you asking questions over your own collections inside the app. A Bot is the same grounded question answering packaged for others: it has fixed knowledge, instructions, and a model, and talks to people through channels like Telegram or a website widget.

Yes, if you connect sources. Monitored channels, playlists, and websites are checked on a schedule; new content is imported automatically and can be rolled into recurring briefings. Static uploads stay as they are until you add more.

Yes. Keldura exposes knowledge folders over MCP so AI agents can query them, and a REST developer API with API keys offers the same programmatically: ask questions, fetch documents, and list citations.

Keldura is a web app that works in any modern browser and installs as a progressive web app on phones and desktops, so it behaves like a native app without an app store download.

Bot conversations draw on your credit balance like your own chats. The daily allowance covers light use; for public-facing bots a paid plan with a larger allowance is the practical choice. The live credit price list shows exact costs per action.

Transcription and indexing run in the background and large imports take time, roughly in proportion to the amount of media. The progress tray shows live status, and anything that fails lands on the Needs attention page with a retry button.

Keldura is in active development: features ship frequently and occasional rough edges are possible. The beta expectations page describes what is stable, what is evolving, and how to report problems.