Use Obscurify.ai models in Cursor

Cursor still lets you override the OpenAI base URL, so you can send its OpenAI-model chat to Obscurify.ai and use an uncensored model. But the override is narrow and has broken other features in the past, so this page is honest about what works and what doesn’t, and shows the MCP route as the robust alternative.

Config verified against Cursor’s settings and community reports on 2026-11-08. The override path is fragile by Cursor’s own design; the MCP route below is tested in our MCP docs.

The base URL override (works, with caveats)

In Cursor, open Settings → Models, scroll to API Keys, and under OpenAI API Key:

  1. Paste your Obscurify.ai key (sk-obs1_...) and turn on Use OpenAI API Key.
  2. Turn on Override OpenAI Base URL and set it to https://obscurify.ai/v1.
  3. Add a custom model whose id is an Obscurify.ai model, e.g. z-ai/glm-5.3-flash-uncensored (Pro) or pqhaz/deepseek-v4.1-flash-1017-heretic (any funded account), and verify it.

Know the limits before you rely on it:

  • It only reroutes the OpenAI-model path. Tab completion, Cursor’s agent indexing and several server-side features still go through Cursor’s own backend, not your endpoint.
  • It can break other providers. With the override on, Cursor’s built-in Claude and other models often return errors until you turn it back off.
  • Verification can be strict. Cursor may reject a custom model id it can’t verify against a known provider.

Get your key: Create an account (a random 16-digit number, no email), add credit on the account page, then click Show API Key. New accounts may get $1 of starter credit while this month’s promo budget lasts. Uncensored models need the Pro plan; decensored models like pqhaz/deepseek-v4.1-flash-1017-heretic work on any funded account.

The MCP route (recommended)

Instead of replacing Cursor’s model, add Obscurify.ai as an MCP server and call its models as tools from inside Cursor’s own agent. This keeps Tab, indexing and your usual model working, and gives Cursor a way to reach an uncensored model when it needs one. See @obscurify/mcp for the one-line setup.

If something fails

  • 402: no balance — add credit on the account page.
  • 403 model_not_on_plan: uncensored models need Pro; decensored models work on any funded account.
  • 404 / verification errors: check the id against /v1/models and keep its prefix. If Cursor won’t verify a custom model, use the MCP route.
  • Other models suddenly erroring: turn Override OpenAI Base URL back off.

What stays private

Prompts and responses aren’t stored on Obscurify.ai’s servers; models with the ZDR tag run only on zero-data-retention routes. Providers see the prompt text but not who sent it. See the privacy policy.

More agents: Obscurify.ai for agents.