DevCycle MCP Getting Started

The DevCycle Model Context Protocol (MCP) Server is based on the DevCycle CLI and enables AI coding tools like Claude Code and Cursor, 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

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

Configure Your AI Client

Claude Code

Step 1: Add DevCycle MCP Server

Run the following command in your terminal:

claude mcp add --transport http devcycle https://mcp.devcycle.com/mcp

Step 2: Manage MCP Connection

Start Claude Code and enter the MCP management interface:

/mcp

Step 3: Authentication

You'll see the DevCycle server listed as "Needs authentication":

1. Select the DevCycle server and press Enter to authenticate
2. This will open a browser page at mcp.devcycle.com for authorization
3. Review and click "Allow Access" to grant permissions
4. If you have multiple organizations, select your desired organization at auth.devcycle.com
5. Return to Claude Code where the server will show as connected

For more details, see the Claude Code MCP documentation.

Claude Desktop

Claude Desktop connects to the DevCycle MCP as a custom connector, so there's no configuration file to edit. Connectors added here are also available on claude.ai.

Step 1: Add the DevCycle Connector

1. Open Claude Desktop and go to Customize → Connectors
2. Click "+ Add", then "Add custom connector"
3. Enter DevCycle as the name and https://mcp.devcycle.com/mcp as the remote MCP server URL, then click "Continue"
4. Keep the default authentication settings and click "Add"

Team and Enterprise plans: An Owner must first add the connector in Organization settings → Connectors using the same URL. Members then find DevCycle under Customize → Connectors and click "Connect".

Step 2: Authentication

1. Click "Connect" on the DevCycle connector if you aren't prompted to sign in automatically
2. This will open a browser page at mcp.devcycle.com for authorization
3. Review and click "Allow Access" to grant permissions
4. If you have multiple organizations, select your desired organization at auth.devcycle.com
5. Return to Claude Desktop and enable DevCycle from the "+" menu → "Connectors" in a conversation

Free plans are limited to one custom connector. For more details, see the Claude custom connectors documentation.

Codex CLI

Step 1: Access MCP Configuration

Locate and edit your Codex configuration file. This file is shared by the Codex CLI, the Codex IDE extension, and the ChatGPT desktop app:

• All platforms: ~/.codex/config.toml

Step 2: Add DevCycle MCP Server

Add the following TOML configuration to enable the DevCycle MCP server:

[mcp_servers.devcycle]
url = "https://mcp.devcycle.com/mcp"

Step 3: Authentication

Run the following command to log in to the DevCycle MCP server:

codex mcp login devcycle

1. This will open a browser page at mcp.devcycle.com for authorization
2. Review and click "Allow Access" to grant permissions
3. If you have multiple organizations, select your desired organization at auth.devcycle.com
4. Start a new Codex session where the DevCycle MCP tools will be active

For more details, see the OpenAI Codex MCP documentation.

Codex App

The Codex app is now part of the ChatGPT desktop app, and shares its MCP configuration with the Codex CLI and IDE extension.

Step 1: Add DevCycle MCP Server

1. Open the ChatGPT desktop app and go to Settings → MCP servers
2. Click "Add server" and choose Streamable HTTP
3. Enter devcycle as the name and https://mcp.devcycle.com/mcp as the URL
4. Save the server, then click "Restart"

Step 2: Authentication

1. Click "Authenticate" next to the DevCycle server in the MCP servers list
2. This will open a browser page at mcp.devcycle.com for authorization
3. Review and click "Allow Access" to grant permissions
4. If you have multiple organizations, select your desired organization at auth.devcycle.com
5. Return to the app where the DevCycle MCP tools will be active in Codex

For more details, see the OpenAI Codex MCP documentation.

Cursor

📦 Install in Cursor

To open Cursor and automatically add the DevCycle MCP, click the install button above. Alternatively, go to Customize → MCPs in Cursor and click "New MCP Server", which opens your ~/.cursor/mcp.json file, then add the following configuration. To learn more, see the Cursor documentation.

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

Authentication in Cursor:

1. After configuration, DevCycle MCP will be listed under Needs Attention as "Needs authentication" in Customize → MCPs
2. Click "Authenticate" 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

VS Code

📦 Install in VS Code

To open VS Code and automatically add the DevCycle MCP, click the install button above. Alternatively, add the following to your workspace .vscode/mcp.json file, or run "MCP: Open User Configuration" from the Command Palette to add it for all workspaces. To learn more, see the VS Code MCP documentation.

{
  "servers": {
    "devcycle": {
      "type": "http",
      "url": "https://mcp.devcycle.com/mcp"
    }
  }
}

Authentication in VS Code:

1. After configuration, run "MCP: List Servers" from the Command Palette, or use the Start code lens in mcp.json
2. Select the DevCycle MCP server and start it
3. VS Code will show a dialog saying the MCP server definition wants to authenticate to mcp.devcycle.com
4. Click "Allow" to proceed with authentication
5. This opens a browser authorization page at mcp.devcycle.com
6. Review and click "Allow Access" to grant permissions
7. If you have multiple organizations, select your desired organization at auth.devcycle.com
8. You'll be redirected back to VS Code with the server now active

OpenCode

Step 1: Add DevCycle MCP Server

Run the following command and follow the interactive prompts to add the DevCycle MCP server (name: devcycle, type: remote, url: https://mcp.devcycle.com/mcp):

opencode mcp add

Step 2: Authenticate

Run the following command to authenticate with DevCycle:

opencode mcp auth devcycle

1. This will open a browser page at mcp.devcycle.com for authorization
2. Review and click "Allow Access" to grant permissions
3. If you have multiple organizations, select your desired organization at auth.devcycle.com
4. Return to your terminal where authentication will complete

The next time you start OpenCode, the DevCycle MCP tools will be available.

For more details, see the OpenCode MCP documentation.

Antigravity CLI

Antigravity CLI replaces Gemini CLI. The same MCP configuration is shared by Antigravity CLI and the Antigravity desktop app.

Step 1: Access MCP Configuration

Locate and edit your Antigravity MCP configuration file:

• All platforms: ~/.gemini/config/mcp_config.json

Step 2: Add DevCycle MCP Server

Add or merge the following configuration to enable the DevCycle MCP server:

{
  "mcpServers": {
    "devcycle": {
      "serverUrl": "https://mcp.devcycle.com/mcp"
    }
  }
}

Step 3: Manage MCP Connection

Start Antigravity CLI with agy, then enter the MCP management interface:

/mcp

Step 4: Authentication

1. Use the arrow keys to select the DevCycle server, choose Authenticate, and press Enter
2. This will open a browser page at mcp.devcycle.com for authorization
3. Review and click "Allow Access" to grant permissions
4. If you have multiple organizations, select your desired organization at auth.devcycle.com
5. If prompted, paste the authorization code back into the terminal
6. Return to Antigravity CLI where the DevCycle MCP tools will be active

For more details, see the Antigravity MCP documentation.

Devin Desktop

Devin Desktop (formerly Windsurf) configures MCP servers for its default Devin Local agent in the Devin config files.

Step 1: Access MCP Configuration

Locate and edit your Devin MCP configuration file:

• macOS/Linux: ~/.config/devin/mcp_config.json
• Windows: %APPDATA%\devin\mcp_config.json

Step 2: Add DevCycle MCP Server

Add or merge the following configuration to enable the DevCycle MCP server:

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

Step 3: Authentication

Run the following command to log in to the DevCycle MCP server:

devin mcp login devcycle

1. This will open a browser page at mcp.devcycle.com for authorization
2. Review and click "Allow Access" to grant permissions
3. If you have multiple organizations, select your desired organization at auth.devcycle.com
4. Return to Devin Desktop and open a new agent tab where the DevCycle MCP tools will be active

For more details, see the Devin MCP documentation.

Available Tools

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

Category	Tools	Description
Feature Management	list_features, create_feature, update_feature, update_feature_status, delete_feature, cleanup_feature, get_feature_audit_log_history	Create and manage feature flags
Variable Management	list_variables, create_variable, update_variable, delete_variable	Manage feature variables
Project Management	get_current_project, select_project	Project selection and details
Self-Targeting & Overrides	get_self_targeting_identity, update_self_targeting_identity, list_self_targeting_overrides, set_self_targeting_override, clear_feature_self_targeting_overrides	Testing and overrides
Results & Analytics	get_feature_total_evaluations, get_project_total_evaluations	Usage analytics
SDK Installation	install_devcycle_sdk	SDK 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
• CLI Reference - Learn about the underlying CLI commands

Getting Help

• GitHub Issues: GitHub Issues
• General Documentation: DevCycle Docs
• Support: Contact Support