Integration · LaunchDarkly

Three MCP servers across feature mgmt, AI Configs, observability.

LaunchDarkly is the largest pure-play feature flag platform. Three hosted MCP servers cover feature management, AI Configs (LLM rollout settings), and observability.

Use the hosted MCPs via OAuth, or self-host locally with an API key. The Federal environment is supported via a --server-url flag, so government and regulated customers get the same agent surface as commercial. Customers include Netflix, Atlassian, Square, IBM.

Official MCPThree product-area servers can be enabled independently. Companion agent-skills repository at skills.sh.

What you can do via MCP

Example prompts the agent runs.

  • “Create a feature flag for the new-checkout feature, default off.”

    Creates the flag in the LaunchDarkly project, sets the default off, returns the SDK snippet for the chosen language.

  • “Toggle the pricing-banner flag on in production for 50% of users.”

    Updates the production targeting rules to a 50% rollout and confirms the change took effect.

  • “List all stale flags across all environments and create a cleanup plan.”

    Pulls flag list, scores by last evaluation, surfaces a prioritized cleanup plan grouped by owner team.

  • “Update targeting rules for the experimental-feature flag to include only beta-testers segment.”

    Edits the targeting rule list with the new segment, surfaces the diff, and applies it to the right environment.

  • “Show me the status of the new-onboarding flag across dev, staging, and production.”

    Returns a per-environment status matrix with rollout percentages and the last change author.

LaunchDarkly · MCP

Local server invocation against the feature management MCP. Hosted alternatives via OAuth are equivalent and skip the env-var step.

LaunchDarkly MCP
docs
# LaunchDarkly MCP, local
# Run: LD_API_KEY=api-abc \
#      npx @launchdarkly/mcp-server

mcp.launchdarkly.create_feature_flag({
  project_key: "default",
  key: "new-checkout",
  name: "New checkout flow",
  variations: [
    { value: true, name: "on" },
    { value: false, name: "off" }
  ],
  defaults: { onVariation: 0, offVariation: 1 },
  temporary: false
})

# Hosted alternative: OAuth via skills.sh agent-skills
# Federal: pass --server-url <gov-endpoint>
One command sample showing how the agent talks to LaunchDarkly. The MCP exposes the platform's primitives as tools; the agent translates the prompt into the right call.

MCP integration

LaunchDarkly MCP server.

Server
Hosted (recommended) at LaunchDarkly endpoints, or local via @launchdarkly/mcp-server
Auth
OAuth (hosted) or LD_API_KEY (local)
Hosting
Hosted (recommended) or local
  • Three product-area servers: feature management, AI Configs, observability
  • AI Configs MCP for managing LLM rollout settings (prompts, model versions, fallbacks)
  • Federal environment supported via --server-url flag
  • Companion agent-skills repo at skills.sh for prebuilt prompts

LaunchDarkly MCP docs

Visual demonstration

What this looks like in practice.

LaunchDarkly/Feature flag
Release

new-checkout

New checkout flow with single-step billing

Environment

production

On50% rollout

Targeting rules

  • segment IS beta-testers ENABLE 100%
  • segment IS internal-users ENABLE 100%
  • ELSE rollout 50%
LaunchDarkly flag state managed via the feature management MCP. Targeting rules and rollout percentage editable through the agent.
LaunchDarkly· Stale flag audit
2 stale of 4

  • old-search-v1

    Stale

    Last evaluated 94 days ago · @search-team

    Pre-redesign search; new search is in production

    Remove

  • checkout-experiment-q3

    Stale

    Last evaluated 61 days ago · @growth

    Experiment ended, ship variant on; flag still gating

    Remove

  • new-onboarding

    Active

    Last evaluated 3 minutes ago · @onboarding

    Active rollout at 50%

    Keep

  • ai-recommendations

    Active

    Last evaluated 1 minute ago · @ai-platform

    AI Config: model A vs model B routing

    Keep
Cross-environment stale flag audit run by the agent. The AI Config flag stays; the post-experiment flag is queued for removal.

CLI alternative

REST API plus MCP.

LaunchDarkly has a comprehensive REST API for direct programmatic access. The MCP is the AI-native interface; the REST API covers CI integration, scripted reporting, and any non-agent workflow. Most teams use both: MCP for agent-driven flag work, REST for build-time and ops automation.

Pairs with these skills

The PM-experimentation skill suite.

This integration pairs with the forthcoming feature-flagging, experiment-design skills. The skill landing pages and SKILL.md sources land in subsequent dispatches; cross-link hyperlinks are added when the skill pages ship.

Cost considerations

LaunchDarkly's MCP is included with the platform subscription.

Subscription included

No per-call cost; usage is bounded by the plan tier you already pay for. Lean on the MCP for exploratory work the way you would lean on the platform UI: the marginal cost of an extra call is zero, and the agent reads or writes the same data your team already has access to.

LaunchDarkly plan covers MCP usage. Federal and enterprise tiers supported for regulated environments.

Recommended patterns

  • Treat the MCP as a thin layer over the platform; permissions cascade from platform-side roles, not a separate MCP credential.
  • Plan-tier limits (seats, projects, environments, event volume) still apply; design agent workflows to live within them.
  • Audit the agent's writes the way you would audit any teammate's: subscription-included does not mean the changes are reversible.

Continue reading.