Anthropic Adds Remote MCP Server Support to Claude Code, Expanding AI Development Capabilities

Jun 16, 2025

Anthropic has expanded its AI coding assistant, Claude Code, with support for remote Model Context Protocol (MCP) servers. This update allows developers to connect directly to third-party tools like Sentry and Linear without managing local infrastructure, using a simple OAuth flow. The move leverages the open MCP standard, which is also being adopted by Google and OpenAI, positioning Anthropic as a key infrastructure provider in the AI ecosystem.

MCP server setup

▪Remote MCP servers are managed by vendors, which eliminates the need for developers to handle infrastructure, deployment, and scaling.
▪The addition of remote MCP support was driven by developer requests for easier customization and access to more information.
▪To connect to a remote MCP server, developers add a vendor URL to their Claude Code configuration and authenticate using an OAuth flow.
▪The new remote server support is available to all Claude Code users, including subscribers to the $20/month Pro plan.
▪Anthropic added remote MCP server support for its AI assistant, Claude Code, on June 18, 2025.

Model Context Protocol standard

▪The Model Context Protocol (MCP) is an open standard introduced by Anthropic in late 2024 for connecting AI models to external tools and data sources.
▪MCP uses JSON-RPC 2.0 for message transport and is designed to solve the "N×M" integration problem where every AI model and tool required a custom connector.
▪Using MCP integrations can increase a developer's Claude API costs by 25-30% due to additional context being pulled from external sources.
▪The introduction of remote MCP support strategically positions Anthropic as an infrastructure provider for AI-tool communication, not just a model vendor.
▪Anthropic's launch partners for remote MCP servers include error tracking service Sentry and project management tool Linear.

Security authentication practices

▪All communication with remote MCP servers must use HTTPS, and servers need to implement proper CORS headers for browser-based clients.
▪The technical implementation for remote MCP requires HTTP transport with mandatory OAuth 2.1 compliance for authentication.
▪Claude Code maintains an approval-required approach for executing actions via MCP and provides audit logs for all operations.

Multi-LLM integration support

▪Other AI coding tools like Cursor and Windsurf also support tool usage over the MCP standard.
▪MCP is designed to be vendor-neutral, allowing a single server to work with any compatible AI system, not just Claude.
▪The MCP standard has been adopted by other major AI companies, including Microsoft, OpenAI, and Google.

Debugging troubleshooting methods

▪Remote MCP introduces network dependencies, meaning that internet connectivity issues or vendor outages can disrupt developer workflows.
▪Common troubleshooting methods for MCP include checking configuration files, verifying environment paths, enabling verbose logging, and restarting the Claude Desktop application.
▪A known issue exists with accessing streamable, OAuth-protected remote servers, making `/see` endpoints a more reliable option.