@featureflip/mcp

MCP (Model Context Protocol) server for Featureflip — manage feature flags from AI coding assistants (Claude Code, Cursor, Copilot, Cline) and autonomous agents.

Setup

Create an API token in the Featureflip dashboard:
- Personal token (ffp_...): Settings → API Tokens — acts as you, for interactive editor use.
- Service token (ffs_...): Organization Settings → Service Tokens — scoped machine identity for CI/agents.
Add the server to your MCP client:

Claude Code

claude mcp add featureflip -e FEATUREFLIP_TOKEN=ffp_your_token -- npx -y @featureflip/mcp

Cursor / Cline / generic JSON config

{
  "mcpServers": {
    "featureflip": {
      "command": "npx",
      "args": ["-y", "@featureflip/mcp"],
      "env": { "FEATUREFLIP_TOKEN": "ffp_your_token" }
    }
  }
}

Configuration

Env var	Required	Default	Purpose
`FEATUREFLIP_TOKEN`	yes	—	`ffp_`/`ffs_` API token
`FEATUREFLIP_API_URL`	no	`https://api.featureflip.io`	API base URL
`FEATUREFLIP_ORG`	no	auto	Org slug (needed only for multi-org personal tokens)

Tools

CRUD: list_projects, list_environments, list_flags, get_flag, create_flag, update_flag, delete_flag, archive_flag, restore_flag, toggle_flag, update_flag_environment_config, get_targeting, update_targeting, manage_variation, list_segments, get_segment

Workflows: flag_status (cross-environment view), find_stale_flags (cleanup candidates), wrap_feature (create flag + get the SDK snippet for your language)

Full reference: https://featureflip.io/docs/integrations/mcp/

Notes

Flag evaluation is not exposed over MCP — use the language SDKs in application code.
Every mutation is audit-logged and attributed to the token.

License

Apache-2.0