LaunchDarkly's skills and hosted MCP server: create, target, roll out and clean up feature flags, and manage AgentControl configs.
by launchdarkly· Developer toolsUnclaimed · Yours? Claim it
4 images
About
LaunchDarkly's own plugin for coding agents. It bundles LaunchDarkly's agent skills, step-by-step playbooks for common feature flag and AgentControl work, with LaunchDarkly's hosted MCP server, so your agent can read and change your flags directly. The skills keep the agent calling the right tools in the right order: a flag cleanup checks removal readiness and code references before it touches any code. What you can do with it - Find stale or launched flags and check whether each one is safe to remove - Create feature flags that fit your codebase's existing patterns - Turn flags on or off, roll out to a percentage of users, and change targeting rules across environments - Remove a flag from code, keep the winning variation and open a pull request, then archive the flag once the change ships - Choose and create metrics, and add tracking calls to your code - Create and update AgentControl configs, with their variations, targeting, tools and LLM-as-a-judge evaluations, organized in projects - Turn on Release Autopilot auto-flagging and auto-releasing for your GitHub repositories How to install it In Claude Code, run /plugin marketplace add launchdarkly/ai-tooling, then /plugin install launchdarkly@launchdarkly-ai-tooling. In Cursor, find LaunchDarkly under Settings > Plugins. The skills alone install with npx skills add launchdarkly/ai-tooling. How it connects The plugin sets up LaunchDarkly's hosted MCP server at https://mcp.launchdarkly.com/mcp/launchdarkly. You sign in to LaunchDarkly with OAuth, so no API key is needed. Access refreshes on its own, and you authorize it again every 90 days. The server acts with the permissions of the member who signed in. Good to know - The hosted MCP server isn't available in LaunchDarkly's federal or EU environments; LaunchDarkly offers a local MCP server for those. - To keep your agent to reading, sign in as a member with a Reader role. Filtering tools in the server address doesn't limit what the connection can do. - When an environment requires approvals, the skills create an approval request instead of changing the flag directly. - Flag data, and anything people wrote into it, goes to your agent's model provider, so review changes before your agent makes them.
Once it’s added
“Which feature flags are stale and should be cleaned up?”
“Create a feature flag for the new checkout flow.”
“Roll out dark-mode to 25% of users in production.”
Before you add it
Reads only
Lists your flags and finds stale ones · Checks a flag's status and removal readiness across environments · Reads AgentControl configs, metrics and error groups
Makes changes · your agent can ask first
Creates flags and changes their targeting, rollouts and on/off state · Removes flags from code in a pull request, then archives them · Creates and updates AgentControl configs and metrics · Changes Release Autopilot settings, after you confirm
Each with its own page
From people who use it
Step by step
Nothing new reaches your agent until a person has looked at it.
Plugins people install alongside LaunchDarkly.
No reviews yet. Used LaunchDarkly? Say how it went.
Write a reviewRuns on your machine.
Runs on your machine.
/plugin marketplace add launchdarkly/ai-tooling /plugin install launchdarkly@launchdarkly-ai-tooling
Scanned and passed · checked 8 h ago
Plugins people install alongside LaunchDarkly.
Matt Pocock's everyday engineering skills: grilling sessions, specs and tickets, TDD, bug diagnosis, code review and writing for agents.
Addy Osmani's engineering workflows for coding agents: spec, plan, build, test, review and ship, with quality gates at each step.
Nobody has shared how they use LaunchDarkly yet.
“Remove the new-checkout-flow feature flag from this codebase.”
“Update the targeting rules for example-feature so it's only enabled for users in Canada.”
“Show me error groups from the last 24 hours.”
Scanned on every change, and again when our rules improve.
Claude CodeSupported
OpenClawSupportedTell us and a person on the trust team will look at it. Or email [email protected].
Security scan
Passed
Safe to install in every agent it supports. Claude plugins.
