LaunchDarkly MCP Server: Let AI Manage Feature Flags

Why Give AI Access to LaunchDarkly?

LaunchDarkly is the feature management platform teams use to control rollouts, run experiments, and kill-switch production issues. When you need to check if a flag is on for a specific segment, review targeting rules, or see which flags changed recently, you switch to the LaunchDarkly dashboard and dig through projects and environments.

An MCP server for LaunchDarkly lets your AI agent query flag states, check targeting rules, list recent changes, and review rollout percentages directly from Claude, Cursor, or Windsurf.

What Becomes Possible

Once your AI agent has LaunchDarkly access:

• "Is the new-checkout-flow flag enabled in production?"
• "What targeting rules are set on the payments-v2 flag?"
• "Which flags were toggled in the last 24 hours?"
• "What percentage of users see the redesigned dashboard?"
• "Are there any flags still targeting individual user IDs?"
• "List all flags in the mobile project that are off in staging but on in prod"

How It Works

1. Go to DataFaucet and enter your LaunchDarkly URL (e.g. https://app.launchdarkly.com)
2. Log into your LaunchDarkly account
3. Browse projects, flags, segments, and audit logs as you normally would
4. Each action becomes an MCP tool your AI can call
5. Deploy and connect to Claude, Cursor, or Windsurf

Works with LaunchDarkly cloud. The AI agent gets read access to whatever your API token permits.

Security Best Practices

• Read-only API token — use a token with Reader role, no flag modification permissions
• Environment-scoped — restrict to specific environments (production read, staging read)
• No SDK key exposure — never grant access to client-side or mobile SDK keys
• Audit log monitoring — LaunchDarkly logs all API access for compliance
• Token rotation — regenerate tokens on a regular schedule

Use Cases by Role

Backend developers: Pre-deploy checks. "Is the flag I depend on enabled in staging?" before shipping code.

Release managers: Rollout oversight. "What's the rollout percentage for the new onboarding flow?" without dashboard navigation.

SREs: Incident response. "Which flags changed in the last hour?" when investigating a production issue.

Product managers: Experiment status. "What segments are targeted by the pricing-test flag?" for experiment reviews.

LaunchDarkly UI vs AI Agent

Getting Started

Create a read-only API token for AI access:

1. Go to Account Settings → Authorization → Access tokens
2. Create a new token with Reader role
3. Scope to specific projects if needed
4. Copy the token

Then point DataFaucet at your LaunchDarkly instance, authenticate with the API token, and your AI agent can query flag states and targeting rules as MCP tools.