Configuring MCP Server with Claude Code

Overview

Claude Code is Anthropic's AI-powered coding assistant that runs in your terminal. It provides native support for Model Context Protocol (MCP) servers through its built-in CLI commands. This article shows you how to connect your Cequence AI Gateway MCP server to Claude Code.

CLI-Based Configuration

Unlike other clients, Claude Code uses CLI commands for MCP configuration - no manual file editing required!

Configuration Methods

Claude Code supports two transport types for MCP servers:

• HTTP: Standard HTTP-based communication (recommended)
• SSE: Server-Sent Events for real-time streaming

Method 1: Quick Setup with NPX (Recommended)

The fastest way to configure your MCP server:

npx @cequenceai/mcp-cli@latest claudecode --url "your-mcp-server-url" --name "your-server-name"

Options

Flag	Description	Required
-u, --url	Cequence MCP server URL	Yes
-n, --name	Server name (default: "cequence-mcp")	No
-t, --transport	Transport type: http or sse (default: http)	No
-k, --api-key	API key for authentication	No
--strict-mcp-config [path]	Write a dedicated config file (default: .mcp-strict.json) for use with strict mode. Optionally pass a target path or directory.	No

Examples

HTTP transport (default):

npx @cequenceai/mcp-cli@latest claudecode --url "https://mcp.aigateway.cequence.ai/v1/servers/abc123" --name "salesforce"

SSE transport:

npx @cequenceai/mcp-cli@latest claudecode --url "https://mcp.aigateway.cequence.ai/v1/servers/abc123" --name "salesforce" --transport sse

Method 2: Direct Claude CLI

Use the Claude CLI directly to add MCP servers:

HTTP Transport

claude mcp add --transport http <server-name> <server-url>

Example:

claude mcp add --transport http salesforce-crm "https://mcp.aigateway.cequence.ai/v1/servers/sf-abc123"

SSE Transport

claude mcp add --transport sse <server-name> <server-url>

Example:

claude mcp add --transport sse github-realtime "https://mcp.aigateway.cequence.ai/v1/servers/gh-xyz789"

Method 3: Strict MCP Config (Isolated Session)

Strict mode makes Claude Code load only the servers from a config file you pass explicitly, ignoring every other MCP configuration source — the project .mcp.json, user settings, plugins, and claude.ai connectors. This is useful when you want a clean, reproducible session that exposes just your Cequence AI Gateway server.

Generate the config with NPX

npx @cequenceai/mcp-cli@latest claudecode --url "your-mcp-server-url" --name "your-server-name" --strict-mcp-config

This writes a .mcp-strict.json file to your current directory:

{
  "mcpServers": {
    "your-server-name": {
      "type": "http",
      "url": "your-mcp-server-url"
    }
  }
}

You can also pass an explicit path or directory:

npx @cequenceai/mcp-cli@latest claudecode --url "your-mcp-server-url" --name "salesforce" --strict-mcp-config ./config

Launch Claude Code in strict mode

claude --mcp-config .mcp-strict.json --strict-mcp-config

The --strict-mcp-config flag tells Claude Code to use only the MCP servers from the file passed with --mcp-config, ignoring all other MCP configuration sources.

Reproducible sessions

Because the strict config file is a plain JSON file, you can commit it to a repository (or share it) so teammates launch Claude Code with the exact same MCP server set every time.

Configuration Parameters

Replace the following placeholders:

• <server-name>: A descriptive name for your MCP server (for example, "salesforce-crm", "github-repo", "jira-project")
• <server-url>: The URL provided by Cequence AI Gateway after creating your MCP server

Naming Conventions

Server names should be:

• Lowercase
• Hyphen-separated (no spaces)
• Descriptive of the service

Good examples:

• salesforce-sales
• github-devops
• jira-project-tracker

Avoid:

• Salesforce Sales (spaces)
• sf (too vague)
• my_server_1 (underscores)

Multiple MCP Servers

Add multiple servers by running the command for each:

# Add GitHub server
claude mcp add --transport http github "https://mcp.aigateway.cequence.ai/v1/servers/github-abc"

# Add Jira server
claude mcp add --transport http jira "https://mcp.aigateway.cequence.ai/v1/servers/jira-xyz"

# Add Salesforce server
claude mcp add --transport http salesforce "https://mcp.aigateway.cequence.ai/v1/servers/sf-123"

Managing MCP Servers

List Configured Servers

claude mcp list

Remove a Server

claude mcp remove <server-name>

Example:

claude mcp remove salesforce-crm

Transport Types

HTTP Transport

• Use case: Standard request/response operations
• Best for: Most API integrations, CRUD operations
• Reliability: Stable, well-tested

claude mcp add --transport http <name> <url>

SSE Transport

• Use case: Real-time streaming, live updates
• Best for: Monitoring, notifications, streaming data
• Features: Persistent connections, push updates

claude mcp add --transport sse <name> <url>

Choosing Transport

Scenario	Recommended Transport
Standard API operations	HTTP
Real-time notifications	SSE
Data queries	HTTP
Live monitoring	SSE
Most integrations	HTTP

Verification

Once configured, verify your MCP server is working:

1. Start Claude Code in your terminal:

   claude

2. List available MCP servers:

   claude mcp list

3. Test with a simple command:

   "List my open Salesforce opportunities"

4. Check that tools from your MCP server are available

Using MCP Servers in Claude Code

In Terminal Chat

Interact with your MCP server through natural language:

> List my open GitHub pull requests
> Create a Jira ticket for the login bug
> Query Salesforce for recent opportunities
> Show me the latest GitLab pipeline status

Common Workflows

Development:

> Show me all failing tests in the CI pipeline
> Create a branch for feature/user-auth
> List merge requests assigned to me

Project Management:

> What Jira tickets are in the current sprint?
> Show overdue tasks in Asana
> List open issues labeled as 'bug'

Data Analysis:

> Query Databricks for user activity metrics
> Get the latest BigQuery results
> Show Elasticsearch logs from last hour

Troubleshooting

Claude CLI Not Found

Issue: claude: command not found

Solutions:

• Install Claude Code: Follow Anthropic's installation guide
• Verify installation: which claude
• Add to PATH if installed but not found

MCP Server Not Connecting

Issue: Server shows as disconnected

Solutions:

• Verify the URL is correct and accessible
• Check your internet connection
• Ensure the MCP server is deployed and running
• Try removing and re-adding the server

Authentication Errors

Issue: "Authentication required" or "Unauthorized"

Solutions:

• Complete OAuth authentication in Cequence AI Gateway
• Verify your credentials are valid
• Re-authenticate the connection

Transport Mismatch

Issue: Server not responding correctly

Solutions:

• Verify you're using the correct transport type
• Most Cequence MCP servers use HTTP transport
• Try switching transport if issues persist

Permission Denied

Issue: "Permission denied" when accessing tools

Solutions:

• Check API permissions in Cequence AI Gateway
• Verify OAuth scopes are sufficient
• Review MCP server endpoint configuration

Tips

1. Use HTTP Transport: Unless you specifically need real-time streaming, use HTTP
2. Descriptive Names: Name servers clearly to remember their purpose
3. One Server Per Service: Configure separate servers for different services
4. Regular Testing: Periodically verify server connections
5. Keep CLI Updated: Update Claude Code regularly for latest features

Prerequisites

Before configuring MCP servers with Claude Code:

1. Claude Code Installed: Install from Anthropic
2. Cequence AI Gateway Access: Create and deploy your MCP server
3. MCP Server URL: Obtain from Cequence AI Gateway after deployment
4. Network Access: Ensure your terminal can reach the MCP server URL

Example Workflow

Complete setup for a development workflow:

# 1. Add GitHub server for code management
claude mcp add --transport http github "https://mcp.aigateway.cequence.ai/v1/servers/gh-abc"

# 2. Add Jira server for issue tracking
claude mcp add --transport http jira "https://mcp.aigateway.cequence.ai/v1/servers/jira-xyz"

# 3. Add Slack server for notifications
claude mcp add --transport http slack "https://mcp.aigateway.cequence.ai/v1/servers/slack-123"

# 4. Verify all servers
claude mcp list

# 5. Start using Claude Code
claude

Then in Claude Code:

> Show my open PRs on GitHub, any related Jira tickets, and post a summary to Slack

Additional Resources

• Claude Code Documentation
• Anthropic Developer Docs
• Model Context Protocol
• Cequence AI Gateway Docs