DevCycle

chính thức

Biến công cụ AI yêu thích của bạn thành trợ lý quản lý tính năng. MCP của DevCycle hoạt động cùng với trợ lý lập trình yêu thích của bạn để bạn có thể tạo và theo dõi các cờ tính năng bằng ngôn ngữ tự nhiên ngay trong quy trình làm việc của mình.

Xem trên devcycle.comDùng thử MCP nàyĐược tài trợ

Bạn có thể làm gì với DevCycle MCP?

  • Manage feature flags — Ask to create, update, or delete flags like "new-checkout-flow" using create_feature, update_feature, or delete_feature.
  • List and inspect projects — Ask to see all features, variables, or switch context with list_features, list_variables, or select_project.
  • Enable targeting and overrides — Ask to turn on targeting for a flag in production via update_feature_status or set self-targeting overrides with set_self_targeting_override.
  • Pull evaluation analytics — Ask for usage metrics like total evaluations for a feature or project using get_feature_total_evaluations or get_project_total_evaluations.
  • Get SDK installation guidance — Ask for setup examples for a specific SDK using install_devcycle_sdk.

Tài liệu

The DevCycle Model Context Protocol (MCP) Server is based on the DevCycle CLI, it enables AI-powered code editors like Cursor and Windsurf, or general-purpose tools like Claude Desktop, to interact directly with your DevCycle projects and make changes on your behalf.

Quick Setup

The DevCycle MCP is hosted so there is no need to set up a local server. We'll walk you through installation and authentication with your preferred AI tools.

Direct Connection: For clients that natively support the MCP specification with OAuth authentication, you can connect directly to our hosted server:

https://mcp.devcycle.com/mcp

Protocol Support: Our MCP server supports both SSE and HTTP Streaming protocols, automatically negotiating the best option based on your client's capabilities.

Alternative Endpoint: If your client has issues with protocol negotiation, use the SSE-only endpoint:

https://mcp.devcycle.com/sse

MCP Registry: If you're using registry.modelcontextprotocol.io, the DevCycle MCP is listed as: com.devcycle/mcp

[!-info] -info info

These instructions use the remote DevCycle MCP server. For installation of the local MCP server, see the reference docs.

Configure Your AI Client

  • Cursor
  • VS Code
  • Claude Code
  • OpenCode
  • Claude Desktop
  • Windsurf
  • Codex CLI
  • Gemini CLI

📦 Install in Cursor

To open Cursor and automatically add the DevCycle MCP, click the install button above. Alternatively, add the following to your ~/.cursor/mcp_settings.json file. To learn more, see the Cursor documentation.

{
  "mcpServers": {
    "DevCycle": {
      "url": "https://mcp.devcycle.com/mcp"
    }
  }
}

Authentication in Cursor:

  1. After configuration, you'll see DevCycle MCP listed as "Needs login" with a yellow indicator
  2. Click on the DevCycle MCP server to initiate the authorization process
  3. This opens a browser authorization page at mcp.devcycle.com
  4. Review and click "Allow Access" to grant permissions
  5. If you have multiple organizations, select your desired organization at auth.devcycle.com
  6. You'll be redirected back to Cursor with the server now active

Available Tools

The DevCycle MCP Server provides comprehensive feature flag management tools organized into 6 categories:

CategoryToolsDescription
Feature Managementlist_features, create_feature, update_feature, update_feature_status, delete_feature, cleanup_feature, get_feature_audit_log_historyCreate and manage feature flags
Variable Managementlist_variables, create_variable, update_variable, delete_variableManage feature variables
Project Managementlist_projects, get_current_project, select_projectProject selection and details
Self-Targeting & Overridesget_self_targeting_identity, update_self_targeting_identity, list_self_targeting_overrides, set_self_targeting_override, clear_feature_self_targeting_overridesTesting and overrides
Results & Analyticsget_feature_total_evaluations, get_project_total_evaluationsUsage analytics
SDK Installationinstall_devcycle_sdkSDK install guides and examples

Try It Out

Once configured, try asking your AI assistant:

  • "Create a new feature flag called 'new-checkout-flow'"
  • "List all features in my project"
  • "Enable targeting for the header-redesign feature in production"
  • "Show me evaluation analytics for the last 7 days"

Next Steps

  • MCP Reference - Complete tool documentation with all parameters
  • - Learn about the underlying CLI commands

Getting Help