Add OPENAI_BASE_URL environment variable support

[tnm]

Summary

Adds support for the OPENAI_BASE_URL environment variable, allowing users to configure custom OpenAI API endpoints for use with:

• Internal enterprise proxies
• API gateways
• OpenAI-compatible providers (OpenRouter, Grok, Together AI, etc.)

This addresses the need for organizations with internal proxy URLs to access OpenAI models through Kit's MCP server.

Changes

• OpenAIConfig.base_url now reads from OPENAI_BASE_URL env var by default
• Works automatically with MCP server and all Kit LLM features
• Added 2 tests for env var configuration
• Documentation updates:
  • README: Brief mention in MCP section
  • MCP docs: Added to Cursor and Claude Desktop examples
  • LLM configuration: New "Enterprise Proxies & Custom Endpoints" section

Usage

export OPENAI_API_KEY="your-key"
export OPENAI_BASE_URL="https://your-proxy.company.com/v1"

Or in MCP config:

{
  "mcpServers": {
    "kit-dev-mcp": {
      "command": "uvx",
      "args": ["--from", "cased-kit", "kit-dev-mcp"],
      "env": {
        "OPENAI_API_KEY": "your-key",
        "OPENAI_BASE_URL": "https://your-proxy.company.com/v1"
      }
    }
  }
}

Test plan

• Unit tests for env var reading
• Existing tests pass
• Manual test with custom endpoint