Claude AI's New Features: What Anthropic Has Been Building

Anthropic has been quietly building features that change how Claude can be used in real workflows. Here's what's actually useful.

Model Context Protocol (MCP)

The biggest infrastructure move Anthropic made in 2025. MCP is an open protocol that lets Claude connect to external tools and data sources.

// MCP server example — Claude can now call your tools
{
  "name": "database-tool",
  "description": "Query the production database",
  "inputSchema": {
    "type": "object",
    "properties": {
      "query": { "type": "string" }
    }
  }
}

What this enables:

• Claude reading your filesystem and editing files
• Claude querying your databases
• Claude using your existing APIs
• Claude integrating with GitHub, Slack, Notion, etc.

This is why Cursor's Claude integration is so powerful — it uses MCP for file access.

Extended Thinking

Claude 3.7 Sonnet has extended thinking — it reasons through complex problems step by step before responding. For hard math, complex coding, and multi-step reasoning tasks, this is meaningfully better.

# Via API
response = anthropic.messages.create(
    model="claude-3-7-sonnet-20250219",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 10000
    },
    messages=[{"role": "user", "content": "Solve this complex problem..."}]
)

Computer Use (Beta)

Claude can control a computer — moving the mouse, clicking, typing, reading screens. It's not perfect, but it works for reliable automation tasks.

Projects

In Claude.ai, Projects lets you give Claude persistent context — your codebase, your writing style, your company docs — across conversations.

This eliminates the "re-explain everything" tax on each new conversation.

What's Coming

Anthropic's research focus is on:

• More capable reasoning (extended thinking improvements)
• Better tool use reliability
• Longer context (> 200K tokens)
• Faster inference at lower cost

The gap between capability and reliability is narrowing faster than the gap between capabilities is widening. Good sign.