What's new on LaunchDarkly

Release automation, feature flags, experimentation & analytics, and AI engineering—on a single platform

changelog
June 11, 2025

Changelog: MCP Server, Observability open beta, and guarded rollouts in Federal

LaunchDarkly MCP server

Image #1

Developers now have more flexibility in how they manage feature flags and AI Configs. By installing our new Model Context Protocol (MCP) server, you can empower AI agents like Cursor, Claude, and Windsurf to interact with your LaunchDarkly data. To get started, read our documentation or follow the tutorial.

Observability now in public Early Access

Image #2

Client-side observability is officially in open beta for every LaunchDarkly account. Every customer now has access to session replay, errors, logs, traces, and alerts. Read the documentation on the in-product observability features and how to set up your SDK.

Guarded rollouts are now available in our Federal environment

Image #3
Guarded rollouts enable teams to gradually release features to a subset of users while monitoring impact and enforcing safeguards, ensuring controlled exposure and minimizing risk. Guarded rollouts are now available to customers using our Federal environment. To learn more about guarded rollouts, read the documentation.

Other improvements

  • We’ve added a new Tutorials section to our documentation site, where you can explore end-to-end examples of how to use LaunchDarkly platform features, courtesy of our Developer Experience team.

  • The LaunchDarkly CLI, ldcli, now includes a command for uploading your sourcemaps, which are then available as part of session replays.

  • We’ve made several improvements to our recent overhaul of individual context targeting for flags based on customer feedback. This includes displaying more rows by default, a bug fix with display names, and more. Thanks to everyone who left feedback so far, we’ve got additional improvements on the way!

  • Left align icons in targeting rule ellipsis menu to be consistent with other menus

Bug fixes

  • Fixed an XSS vulnerability in our login redirect.

  • Fixed a bug with invalid variation options being displayed for some flags when stopping monitoring on a guarded release.

  • Fixed a bug with autocomplete in our JSON targeting editor not working properly in some cases

  • When trying to delete an AI Configs model that doesn’t exist, properly return a 404 instead of a 500

  • Fixed a bug with displaying an incomplete list of available approval reviewers in some views