How integrations work
Every integration shares the same shape: an admin connects it once, members push approved content to it, and delivery is queued, retried, and visible. This page is what is common to all of them; the per-provider guides cover the setup that differs.
- Updated
- 8 October 2026
- Read
- 5 min
- For
- Admins choosing a destination
Destinations differ in what they accept.
Pick by the shape of what you want to send, not by preference.
| Destination | Accepts | Pushing the same thing again | Where it lands |
|---|---|---|---|
| Jira | AcceptsOne finding per issue | Pushing the same thing againUpdates the same issue | Where it landsLive project picker |
| Linear | AcceptsOne finding per issue | Pushing the same thing againUpdates the same issue | Where it landsLive team picker |
| Notion | AcceptsFindings singly or in a batch, recaps, transcripts, reports, cited answers, and visual timelines when ready | Pushing the same thing againRewrites the same page | Where it landsLive page picker |
| Slack | AcceptsFindings singly or in a batch, recaps, cited answers | Pushing the same thing againPosts a new message each time | Where it landsLive channel picker |
| Webhook | AcceptsEverything above, as signed JSON | Pushing the same thing againAlways posts; your receiver decides | Where it landsYour endpoint URL |
Connecting and pushing are different permissions.
- Connecting a destination
- Workspace admin or owner. It is workspace configuration, not a personal setting.
- Choosing where content lands
- Workspace admin or owner, from the same card.
- Pushing content
- Member and above, on a paid plan. Findings must be approved first.
- Retrying a failed delivery
- Member and above.
- Viewers
- Can see destinations and delivery history, and can send nothing out of the workspace through any of them.
Queued, retried, deduplicated, visible.
A push that vanishes silently would be worse than no integration at all.
Queued, never silent
A push is written down before it is sent, so the app says "Queued" rather than pretending it already arrived. Integration activity in Settings shows every delivery and what it created.
Retried automatically
Transient provider failures retry with growing backoff. A revoked credential or a deleted project fails fast instead, with the provider's own reason on the record.
Never duplicated
Pushing unchanged content twice is recognised rather than filed twice. On Jira, Linear, and Notion the existing item is updated in place; the push dialog shows its key and the button becomes "Push again".
Retryable by hand
Failed and given-up deliveries carry a Retry button. Retrying re-queues the same delivery, so it cannot create a second item downstream.
Setup differs per provider: Jira, Linear, Notion, Slack, and webhooks.
Integrations, answered.
Notion, Slack, and the generic webhook. Jira and Linear deliberately take one finding per issue, so every issue in a tracker is one somebody chose to file rather than a bulk dump.
Connecting a destination works on every plan, including free. Pushing content to it requires Pro or above. That split is intentional: a workspace can be set up before it is upgraded.
Recaps, transcripts, reports, and cited Q&A answers, depending on the destination. Visual timelines and frame citations are offered only after that recording has ready visual evidence. Documents and document answers are not pushable yet.
No. A pasted credential is checked with the provider before it is saved, so a typo is caught in the dialog. Beyond that, push something real and watch it in Integration activity.
They stay where they are: they are yours. Citesvue forgets the credential and stops sending. Delivery history and links to created items remain visible.
Not yet. Every push is triggered by a person, and findings are reviewed first. Automatic rules are on the roadmap.