01
Confirm the ChatGPT workspace has custom remote MCP connector support enabled.
Install Guide
Install Enchant in ChatGPT
ChatGPT's remote MCP support is different from a coding CLI, but the underlying requirement is the same: a hosted MCP server with OAuth metadata, protected-resource discovery, and a browser sign-in flow. Enchant now exposes those pieces so it can fit ChatGPT's custom connector path.
Recommended Setup
Start with hosted MCP first.
ChatGPT can connect to remote MCP servers through its custom connector flow when workspace settings allow developer-mode MCP installs.
02
Import the Enchant MCP server URL in ChatGPT's connector settings or app configuration flow.
03
Complete the browser OAuth handoff to your Enchant account.
04
Enable the connector in chat or deep research so ChatGPT can call Enchant tools when needed.
Configuration
Use the exact configuration this client expects.
Server URL to import
ChatGPT should discover OAuth metadata from the hosted MCP server and protected-resource metadata routes.
https://askenchant.com/mcpResponses API MCP tool example
Use this pattern when testing remote MCP servers through the API or deep research tooling.
{
"type": "mcp",
"server_label": "enchant",
"server_url": "https://askenchant.com/mcp",
"require_approval": "never"
}Fallback Option
If your workspace does not expose custom MCP connectors yet
Use Codex, Claude Code, or Cursor first. Enchant's MCP server is ready for hosted OAuth, but ChatGPT availability still depends on workspace settings and rollout state for custom connector installs.
Operational Notes
Small details that make the setup reliable.
01
ChatGPT's MCP support is best for hosted remote servers rather than local stdio installs, which matches Enchant's design.
02
OAuth is the intended auth path here. Unlike coding CLIs, ChatGPT is not the place to rely on a manually pasted bearer token.
03
For API-based testing, remember that MCP tools used in deep research need the approval mode and tool policy you expect.