Connect the LaunchDarkly MCP server to Unblocked to give AI agents live feature flag context

A flag key in the code says little on its own. Whether it is on in production, who it targets, and whether it is safe to remove all live in LaunchDarkly, so an agent editing flagged code either guesses or sends a developer to the dashboard. LaunchDarkly’s MCP server exposes that flag state to AI agents, and Unblocked connects to it once so every agent can ask about a flag without its own LaunchDarkly setup.

This is a third-party MCP server: LaunchDarkly hosts and maintains it, and Unblocked connects to it as a verified MCP server. Nothing is ingested or indexed — Unblocked calls the read-only tools you select at the moment a question needs them, so answers reflect each flag’s targeting and rollout as they stand right now.

Wiring LaunchDarkly’s MCP server straight into an agent tells it how a flag is configured. Unblocked’s Context Engine reasons across those results and your other sources — following a stale flag to the pull request that introduced it, the ticket that scoped the rollout, and the Slack thread where the team agreed to keep it — and returns one cited answer.

Once connected, any MCP-compatible coding agent — Claude Code, Cursor, Devin, GitHub Copilot, Codex — reaches LaunchDarkly context through a single Unblocked MCP server connection.

What you get

• Live flag state across environments — Agents see whether a flag is on, launched, or inactive in each environment, along with its rules, targets, and variations, straight from LaunchDarkly’s own MCP server.
• Safer flag cleanup — Agents can find stale flags and check removal readiness before deleting flagged code, instead of guessing whether a flag still serves traffic.
• Read-only, flag-safe tool control — Only read-only tools are offered, so agents can inspect flags but never toggle, retarget, or archive them. Choose which tools Unblocked may call from Settings → Data Sources.
• Context Engine reasoning — LaunchDarkly results are reconciled with your code, pull requests, tickets, and conversations, so a flag leads to the change and the decision behind it — with citations.

Prerequisites

• An Unblocked account (free to start).
• A LaunchDarkly account outside the federal and EU environments, where the hosted MCP server isn’t available, and a member role that can authorize the server over OAuth.
• An MCP-compatible coding agent (Claude Code, Cursor, Devin, etc.) to use the context in your editor.

Step 1: Connect LaunchDarkly to Unblocked

1. In Unblocked, go to Settings → Data Sources → Connect another data source, and under Cloud Infrastructure select LaunchDarkly.
2. The LaunchDarkly MCP server URL is filled in for you. Choose OAuth, click Connect, and sign in to LaunchDarkly to authorize Unblocked.
3. Once Unblocked discovers the server’s tools, select the read-only tools it may use and click Save Settings.

Step 2: Install the Unblocked MCP server

On macOS or Linux, one command auto-detects and configures the supported agents on your machine:

On Windows, or when you prefer remote MCP, add the server manually using the configuration below, or follow the manual install guide .

claude mcp add -s user --transport http unblocked https://getunblocked.com/api/mcpsse

{
  "mcpServers": {
    "unblocked": {
      "url": "https://getunblocked.com/api/mcpsse"
    }
  }
}

https://getunblocked.com/api/mcpsse

LaunchDarkly hosts and maintains this MCP server. Unblocked calls only the read-only tools you select, at the moment a question needs them:

• list-flagsReturns a filterable, paginated list of the feature flags in one environment.
• get-flagReturns one flag’s configuration in an environment — its rules, individual targets, fallthrough, prerequisites, and variations.
• get-flag-status-across-envsReturns a flag’s lifecycle state in each environment, summarising where it is active, inactive, or launched.
• get-flag-healthReturns a combined health view of one flag, covering lifecycle state, last request time, targeting summary, age, and whether it is temporary.
• find-stale-flagsReturns flags ranked as cleanup candidates by how stale they are.
• check-removal-readinessReturns a safe, caution, or blocked verdict on removing a flag, with a recommended value to hard-code in its place.

To give Claude Code, Cursor, GitHub Copilot, Codex, or another MCP-compatible AI coding agent access to LaunchDarkly context, connect LaunchDarkly to Unblocked and add the single Unblocked MCP server to your agent.

Unblocked acts as the MCP context layer for your LaunchDarkly data, searching feature flags and environments alongside your other connected sources and returning cited results through these tools:

• context_researchResearches a question across your connected sources and calls the LaunchDarkly tools you selected when flag state, targeting, or flag health is relevant, returning a synthesized answer with citations.
• context_get_urlsFetches the full content behind URLs your agent encounters across connected sources — Jira, Linear, Notion, Confluence, Slack, GitHub, and more.

Questions you (or your agent) can ask Unblocked about your LaunchDarkly data:

• Is the new-checkout flag fully launched in production, and which pull request added the code path it guards?
• Which stale flags are safe to remove, and who owns the code that still references them?
• Why does this flag target only internal users in staging, and is there a ticket or Slack thread explaining the rollout plan?
• Before I delete this flagged branch, does LaunchDarkly say the flag is ready for removal?