Skip to main content

​
Overview

LaunchDarkly is a one-click OAuth connector that gives Rootly AI read access to the LaunchDarkly projects your account can see. When an incident correlates with a recent flag change (which they very often do), Rootly AI can pull the flag’s current targeting, its recent audit history, and the exact change that landed just before the incident started — all automatically. Every query is read-only. Rootly AI can inspect flags, targeting rules, and audit logs but cannot toggle flags, change targeting, or modify anything on the LaunchDarkly side.

​
Before You Start

  • A LaunchDarkly account with permission to authorize third-party integrations for your organization.
  • Read access on the LaunchDarkly projects and environments you want Rootly AI to be able to see — Rootly AI inherits the authorizing user’s read scope.

​
Setup

Open Configuration → AI → Connectors and click Connect on the LaunchDarkly card. Sign in to LaunchDarkly and authorize Rootly AI. The connection covers every project and environment the authorizing LaunchDarkly user can read — there’s no per-project scope inside the OAuth flow. When authorization completes, the card flips to Connected.

​
What Rootly AI Can Read

  • Flags — flag names, descriptions, and current state across environments.
  • Targeting — the current targeting rules per flag per environment.
  • Audit logs — the history of flag changes, including who changed what and when.
  • Environments — the list of environments in the project (production, staging, etc.).
Rootly AI reaches for LaunchDarkly hard on the “what changed in the last hour?” question. Feature-flag flips are a top cause of incidents, and having the flip history in the investigation output means responders can rule it in or out immediately.

​
During an Incident

“A feature flag flipped 3 minutes before the spike. Here’s the change.” Rootly AI correlates the incident’s start time against LaunchDarkly’s audit log. When a flag flip lands in the window right before the incident, Rootly AI surfaces the flag name, the exact change (targeting rule diff, rollout percentage, environment), and who flipped it. Feature-flag flips are a top cause of incidents; catching them in the first minute of the investigation cuts triage time meaningfully.

​
Best Practices

  • Include a description on every flag. Rootly AI cites flags by name and description; a well-described flag makes the investigation output actually useful.
  • Use meaningful audit log messages when flipping flags. LaunchDarkly’s audit log captures the comment attached to a change — a good comment (“emergency disable — payment failures on checkout”) means Rootly AI can explain why the flip happened, not just that it happened.
  • Pair with a code connector. LaunchDarkly tells Rootly AI which flags changed; GitHub tells it whether the flag references correspond to code that shipped. Together they’re the strongest “was this deploy vs a flag flip?” signal.

​
Troubleshooting

The authorizing LaunchDarkly user doesn’t have read access to those projects — Rootly AI inherits that user’s LaunchDarkly permissions. Either grant the user read access on the missing projects, or reconnect from a LaunchDarkly account that already has broader coverage.
Either the flag is in an environment or project the connector can’t read, or its name has changed recently and the previous name is being cached in the investigation prompt. Confirm the flag key in LaunchDarkly matches what you’re asking about.
If a LaunchDarkly admin revokes the Rootly AI OAuth grant from LaunchDarkly’s side, Rootly’s card still shows Connected but every query fails. Reconnect from Rootly to reauthorize.

​
Frequently Asked Questions

No. Connectors are read-only. Rootly AI can inspect flag state and history but cannot invoke write actions.
Coverage depends on how your LaunchDarkly instance is reachable. Contact support to confirm compatibility with Federal or dedicated deployments.
Rootly does not maintain a persistent copy of LaunchDarkly data — queries run at investigation time and the response is used in reasoning only. Query results do appear in the LLM traces Rootly logs for quality monitoring, as with every tool call Rootly AI makes. See Data Privacy for Rootly AI for the retention boundary.

Connectors Overview

All connectors and how they fit together.

GitHub

Pair LaunchDarkly with GitHub to answer “was this a flag flip vs a deploy?”

Data Privacy for Rootly AI

What Rootly AI sees, retention, and model training controls.