Docs | MCP Configuration

Ansight Studio exposes a local MCP endpoint on your desktop:

http://127.0.0.1:45125/mcp/

http://localhost:45125/mcp/ also works, but use one form consistently in your client config.

Before You Configure a Client

Launch Ansight Studio on the same machine as your AI client.
Confirm the MCP panel in Studio shows the server as running.
Use the full /mcp/ path in the client configuration.
Restart the client after changing MCP settings if it does not hot-reload server config.

Ansight Studio is a local desktop MCP server. For normal local development, you do not need extra authentication headers.

Codex supports MCP in both the CLI and the IDE extension, and the configuration is shared between them.

Recommended setup:

Add Ansight from the Codex CLI:

codex mcp add ansight --url http://127.0.0.1:45125/mcp/

Verify it is configured:

codex mcp list

Manual config:

You can also add the server directly to ~/.codex/config.toml:

[mcp_servers.ansight]
url = "http://127.0.0.1:45125/mcp/"

Official Codex MCP docs:

Cursor reads MCP configuration from mcp.json.

Global config:

Create ~/.cursor/mcp.json:

{
  "mcpServers": {
    "ansight": {
      "url": "http://127.0.0.1:45125/mcp/"
    }
  }
}

Project config:

If you want the server scoped to one repository, create .cursor/mcp.json in the project:

{
  "mcpServers": {
    "ansight": {
      "url": "http://127.0.0.1:45125/mcp/"
    }
  }
}

For a local desktop service like Ansight, global config is usually the simpler choice.

Official Cursor MCP docs:

Claude Code supports remote HTTP MCP servers and local/project/user configuration scopes.

Recommended setup:

Add Ansight with the Claude CLI:

claude mcp add --transport http ansight http://127.0.0.1:45125/mcp/

Verify it is configured:

claude mcp get ansight

You can also list all configured MCP servers:

claude mcp list

JSON config:

If you prefer JSON-based setup, add the equivalent config:

{
  "mcpServers": {
    "ansight": {
      "type": "http",
      "url": "http://127.0.0.1:45125/mcp/"
    }
  }
}

For a machine-local Studio endpoint, prefer a user or local scope instead of checking workstation-specific loopback config into a shared repo.

Official Claude Code MCP docs:

For IDE-specific MCP setup, use the vendor documentation directly:

Both can connect to the same local Ansight Studio endpoint:

http://127.0.0.1:45125/mcp/

Make sure Ansight Studio is running before you open the AI client.
Use the exact MCP endpoint path, including /mcp/.
Restart the client if the MCP server does not appear after editing config.
If you have multiple MCP servers configured, keep the Ansight server name short and explicit, such as ansight.
Remember that Studio listens locally, so the client must run on the same desktop machine.