Skip to main content
Connic

Retrieval Sources

Connect the places where your team already writes: Notion, Confluence, Superhuman Docs (Coda), or any website. Connic syncs their content into your agents' retrieval on a schedule, so answers stay current without manual uploads.

How retrieval sources work

One pipeline for every source: connect, scope, sync, retrieve.

01

Connect a source

Authorize Notion, Confluence, or Superhuman Docs (Coda) with a scoped token, or point the website crawler at a URL. Connected sources read only what you share with them; the crawler stays inside the boundaries you set.

02

Scope what syncs

Pick spaces, pages, databases, labels, or crawl boundaries. Content outside the scope is never read, and every source syncs into a namespace you choose.

03

Sync on a schedule

From every 15 minutes for API-backed sources, or every 12 hours for website crawls, up to weekly, Connic pulls changed content, converts it to clean text, chunks it, and embeds it for semantic search. Unchanged pages are skipped.

04

Agents retrieve it

Every agent in the project can query the retrieval at run time and answer from the freshest synced version, with namespaces to keep sources separated.

Retrieval sources FAQ

Retrieval sources sync external content into your project's retrieval automatically. Instead of uploading documents by hand, connect Notion, Confluence, Superhuman Docs (Coda), or a website once, and Connic keeps the content chunked, embedded, and retrievable on your schedule.

Uploads are one-time snapshots; a synced source stays current. When someone edits the wiki page or publishes a new doc, the next sync updates the retrieval without anyone re-uploading anything. Both end up in the same retrieval and are retrieved the same way.

Notion, Confluence, and Superhuman Docs (Coda) for internal content, plus a website crawler for anything public like docs or help centers. More sources are on the way.

Each page is converted to clean text, split into chunks, and embedded for semantic search in your project's retrieval, filed under a namespace per source. Agents query it at run time and retrieve the most relevant chunks, no matter which source they came from.

No. Retrieval sources are read-only by design: content is pulled into retrieval, never written back. When an agent should act on an external system, that is what connectors are for.