Connecting an integration
Integrations let Dave read from, and act on, the systems you already run. For the specifics of one provider, see its page under Integrations.
Before you start
Section titled “Before you start”You need the Manage integrations permission, held by the Owner, Infra Owner and Service Infra Owner roles. Developer can view connections but not add, reconnect or disconnect them.
The walkthrough
Section titled “The walkthrough”-
Open Integrations. A new workspace reads Nothing connected yet.
-
Choose Add integration. The catalog groups providers by Source Code, Cloud, Communication, Workflow, Knowledge, Monitoring and Compliance.
-
Pick a provider and complete its flow — see The three flow shapes.
-
Confirm the connection reads Active. Until it does, the flow is not finished.
-
Choose Run health check. Active means “set up correctly”, not “working right now” — a connection can be Active and still fail, usually because the credential was revoked afterwards. The list shows an N failing pill for that.
The three flow shapes
Section titled “The three flow shapes”Straight to a consent popup
Section titled “Straight to a consent popup”Approve the access in the provider’s popup and it closes itself.
Asana · Confluence · GitHub · Jira · Linear · monday.com · Notion · Sentry · Slack
If the popup is blocked, the connection row keeps an Open the setup page action — nothing is lost.
A form, then a consent popup
Section titled “A form, then a consent popup”Azure DevOps asks for your organization name first, then sends you to Microsoft.
A form, then a step in your own console
Section titled “A form, then a step in your own console”These collect credentials or identifiers, then wait on your side. The connection sits at Setup incomplete meanwhile.
| Provider | What finishes it |
|---|---|
| AWS | Deploying a CloudFormation stack |
| Google Cloud | Running a Cloud Shell setup command |
| Kubernetes | Installing the Dave agent with Helm |
| Microsoft Azure | Nothing — Dave validates the credentials |
| Cloudflare | Nothing — Dave validates the token |
| Datadog (API) | Nothing — Dave validates the keys |
| PagerDuty | Nothing — Dave validates the key |
| Vanta | Nothing — Dave validates the credentials |
Most of these also ask for a Label — the name you use to refer to the connection in conversation, as in “use prod to list S3 buckets”. Labels are unique per workspace, up to 64 characters. Kubernetes calls it Cluster name.
Statuses
Section titled “Statuses”| Badge | Meaning | What to do |
|---|---|---|
| Active | Set up correctly and usable. | Nothing. |
| Setup incomplete | Started but never finished — nothing is connected yet. | Resume setup, or Delete to start over. |
| Error | The connection broke. The row carries the reason. | Fix it at the provider, then Connect again. |
| Disconnecting | Teardown started; provider-side resources are still being removed. | Finish any cleanup the provider needs. |
| Disconnected | Fully removed. Grouped separately in the list. | Nothing. |
Connections are sorted worst-first, so anything needing attention floats to the top.
While a provider is being validated Dave polls with a widening interval — 2 seconds, backing off to 15 — so leaving the page open is fine.
Connecting more than one
Section titled “Connecting more than one”Each provider connects per account, workspace, site, cluster or project, and offers Add another (noun):
| Unit | Providers |
|---|---|
| Account | AWS, Cloudflare, GitHub, monday.com, PagerDuty, Vanta |
| Workspace | Asana, Linear, Notion, Slack |
| Organization | Azure DevOps, Datadog (API), Sentry |
| Site | Confluence, Jira |
| Project | Google Cloud |
| Subscription | Microsoft Azure |
| Cluster | Kubernetes |
Atlassian scopes consent per site and Sentry per organization, so those are connected separately rather than added to an existing connection.
Troubleshooting
Section titled “Troubleshooting”| Symptom | Fix |
|---|---|
| Popup was blocked | Use Open the setup page on the connection row. |
| Stuck at Setup incomplete | Resume setup, or Delete to discard and start over. |
| Gave up waiting for the provider | Confirm you ran the setup step, then Resume setup. Scripts re-run safely. |
| Health check fails on Active | Usually a revoked or expired credential — Connect again to re-authorize. |
| Shows Error | Read the reason on the row, fix it at the provider, then Connect again. |
Two providers need more than Connect again: Microsoft Azure needs a fresh client secret first, and Kubernetes has no credential to re-authorize — fix the agent, or disconnect and reinstall.