← Back to Docs

MCP Integration

Manage feature flags directly from your AI assistant. FlagDrop's hosted MCP server gives AI tools full access to your flags, environments, rollouts, and targeting rules — just add a URL and authenticate.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources. Instead of copy-pasting flag configurations or switching between terminals and dashboards, your AI assistant can manage flags directly.

FlagDrop's hosted MCP server exposes 14 tools, giving AI assistants the ability to create flags, toggle rollouts, manage targeting rules, push configs, and query analytics — all through natural language.

Supported Clients

✓

Claude Code

Anthropic CLI for coding with full MCP tool support

✓

Claude Desktop

Anthropic desktop app for conversations with tool access

✓

ChatGPT Desktop

OpenAI desktop app with MCP connector support

✓

Cursor

AI-native code editor with built-in MCP integration

✓

Windsurf

Codeium AI editor with MCP tool support

✓

VS Code + Copilot

GitHub Copilot agent mode with MCP tools

Setup

FlagDrop's MCP server is fully hosted — just add the URL to your AI tool and sign in. No installation, no API keys, no configuration files.

  1. Add https://mcp.flagdrop.io/sse to your MCP client
  2. You'll be redirected to sign in with your FlagDrop account
  3. Authorize the connection — that's it

Configuration by Client

Claude Code

Add to your project's .claude/mcp.json or global ~/.claude/mcp.json:

.claude/mcp.json
{
  "mcpServers": {
    "flagdrop": {
      "url": "https://mcp.flagdrop.io/sse"
    }
  }
}

Claude Desktop

Open Settings → Developer → Edit Config, or edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

claude_desktop_config.json
{
  "mcpServers": {
    "flagdrop": {
      "url": "https://mcp.flagdrop.io/sse"
    }
  }
}

ChatGPT Desktop

Open Settings → Tools → Add MCP Server. Enter the URL:

MCP Server URL
https://mcp.flagdrop.io/sse

ChatGPT will prompt you to authenticate with your FlagDrop account on first use.

Cursor

Open Settings → MCP → Add Server, or edit ~/.cursor/mcp.json:

~/.cursor/mcp.json
{
  "mcpServers": {
    "flagdrop": {
      "url": "https://mcp.flagdrop.io/sse"
    }
  }
}

Windsurf

Open Settings → MCP → Add Server. Enter the URL:

MCP Server URL
https://mcp.flagdrop.io/sse

VS Code + GitHub Copilot

Add to your workspace .vscode/mcp.json:

.vscode/mcp.json
{
  "servers": {
    "flagdrop": {
      "url": "https://mcp.flagdrop.io/sse"
    }
  }
}

Available Tools

Once connected, your AI assistant has access to these tools:

list_projectsList all projects in your organization
list_flagsList all flags in a project with their current state
create_flagCreate a new feature flag (boolean, string, number, or JSON)
update_flagUpdate flag properties (description, default value, scope)
delete_flagDelete a flag (requires confirmation)
toggle_flagEnable or disable a flag in a specific environment
set_rolloutSet rollout percentage for gradual rollouts
add_ruleAdd targeting rules (attribute-based targeting)
push_configManually trigger a config push to cloud storage
get_sync_statusCheck if environments are in sync with latest config
rollbackRoll back to a previous config version
panicEmergency: disable all flags in an environment instantly
scheduleSchedule a flag change for a future time
set_overrideOverride a flag value for child projects

Example Usage

Once configured, just talk to your AI assistant naturally:

AI assistant session
> "Create a boolean flag called dark-mode in the webapp project"

Created flag 'dark-mode' (boolean, default: false) in project 'webapp'

> "Enable dark-mode at 25% rollout in production"

Set dark-mode rollout to 25% in production environment.

> "What flags are enabled in staging?"

3 flags enabled in staging:
  - dark-mode: true (25% rollout)
  - new-checkout: true (100%)
  - beta-api: false

> "Schedule dark-mode to go to 100% tomorrow at 9am"

Scheduled: dark-mode will be set to 100% rollout at 2026-04-07T09:00:00Z

Security

  • Authentication via OAuth — sign in with your FlagDrop account, no API keys to manage
  • Each MCP session is scoped to your organization — you only see your own data
  • All communication is over HTTPS
  • Sessions can be revoked from your FlagDrop account settings

Ready to manage flags with AI?

Get Started Free