Skip to main content
Cursor connects to Neosantara in two ways: using the OpenAI-compatible gateway as your primary coding model provider (Chat, Composer, Agent), and configuring Neosantara as a Model Context Protocol (MCP) server for supplemental tools.

Method 1: Primary Coding Models (OpenAI-Compatible)

Route Cursor Tab, Chat, and Composer requests through Neosantara:
1

Open Cursor Model Settings

Press Ctrl + Shift + J (Windows/Linux) or Cmd + Shift + J (macOS) to open Cursor Settings, then navigate to the Models tab.
2

Configure OpenAI Credentials

  1. Enter your Neosantara API key (nsk_...) in the OpenAI API Key field.
  2. Toggle on Override OpenAI Base URL.
  3. Input the gateway base URL:
3

Add Neosantara Models

Click + Add model and enter the following identifiers:
  • deepseek-v4.1-flash (Recommended for Composer and Agent workflows)
  • gemini-3.8-flash (Recommended for low latency code chat)
  • gpt-5.4-mini (Cost-efficient general tasks)

Method 2: MCP Tool Server

Attach Neosantara tools (model discovery, image generation, web audits) directly to Cursor’s agent environment.
1

Navigate to MCP Settings

Go to Cursor Settings -> Features -> MCP.
2

Add Server Configuration

Click + Add New MCP Server or add this block directly to ~/.cursor/mcp.json:
3

Verify Connection Status

Check the MCP settings tab. Confirm that the neosantara entry displays a green status indicator with available tools (chat, list-models, generate-image).
MCP tool connections require an API key prefixed with nsk_mcp_. Model inference calls in Method 1 use standard gateway API keys (nsk_...).

Next Steps