Dev.to AI 🤖 Ai 👁 0 📖 3 min read

Use any of 398 uncensored models from Cursor, Cline or Claude Code with one key

Most coding agents now take a custom base URL. That means the model behind Cursor, Cline or Claude Code does not have to be the one the tool ships with. If you want an uncensored or abliterated open-weight build instead,

Most coding agents now take a custom base URL. That means the model behind Cursor, Cline or Claude Code does not have to be the one the tool ships with. If you want an uncensored or abliterated open-weight build instead, the switch is a base URL, a key and a model id.

This walks through the three tools with UncensoredLLM, an API that serves 398 abliterated and uncensored models on one prepaid key in both the OpenAI and the Anthropic wire formats. Prices below are from the live catalog on 2026-11-20.

The three values every tool needs

Setting Value
OpenAI-compatible base URL https://api.uncensoredllm.ai/v1
Anthropic-compatible base URL https://api.uncensoredllm.ai (no /v1)
API key sk-ullm-... from the console
Model id any id on the models page

Two ids the guides recommend as a starting pair:

Model id Input $/M Output $/M
glm-5.3-uncensored $0.282 $0.846
glm-5.3-flash-uncensored $0.1053 $0.3685

Cursor

Cursor takes a custom OpenAI key and base URL in its Models settings.

  1. Open Cursor Settings (Cmd/Ctrl + ,) and go to Models.
  2. Turn on OpenAI API Key and paste the key.
  3. Turn on Override OpenAI Base URL, enter https://api.uncensoredllm.ai/v1 and save. The /v1 at the end matters.
  4. Under Models, add the ids you want, exactly as the models page shows them (for example glm-5.3-uncensored, not the display name).

Ask and Agent mode both work. Agent requests carry the conversation, the files Cursor read and the tool results, so input tokens are most of the bill. A build with a low input price costs the least per turn; at the catalog prices, a turn of 40,000 input and 1,200 output tokens is about $0.01 on GLM-5.3 Uncensored and about $0.0047 on GLM-5.3 Flash Uncensored.

Full steps and the common errors: uncensoredllm.ai/guides/cursor.

Cline

Cline's OpenAI Compatible provider takes a base URL, a key and a model id.

  1. Open Cline's settings (the gear in the Cline panel).
  2. Set API Provider to OpenAI Compatible.
  3. Enter Base URL https://api.uncensoredllm.ai/v1 and paste the key as API Key.
  4. Type a Model ID, save, and start a task.

From the CLI, the same four values in one line:

cline auth -p openai-compatible -k "sk-ullm-live-..." -m glm-5.3-uncensored -b https://api.uncensoredllm.ai/v1

Cline also has Use different models for Plan and Act. Plan turns are mostly input and a short answer, so a fast, cheap build does them well; Act turns make the edits and run the commands, so that is where a stronger build earns its price.

Full steps: uncensoredllm.ai/guides/cline.

Claude Code

Every chat model on the API also answers Anthropic's Messages API at /v1/messages, so Claude Code connects directly, with no proxy:

export ANTHROPIC_BASE_URL="https://api.uncensoredllm.ai"
export ANTHROPIC_AUTH_TOKEN="sk-ullm-live-..."
export ANTHROPIC_MODEL="glm-5.3-uncensored"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="glm-5.3-flash-uncensored"
claude

Leave /v1 off the base URL; Claude Code adds the path itself. ANTHROPIC_DEFAULT_HAIKU_MODEL is the model Claude Code uses for background tasks, so a cheap Flash build fits there. /model switches the main model inside a session.

Check the key and the id before you start:

curl -s https://api.uncensoredllm.ai/v1/messages \
  -H "Authorization: Bearer sk-ullm-live-..." \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model": "glm-5.3-uncensored", "max_tokens": 64, "messages": [{"role": "user", "content": "Reply with the single word ok."}]}'

claude-code-router is only needed if you want to mix providers per task. Full steps: uncensoredllm.ai/guides/claude-code.

Picking a build

The catalog has three kinds of build:

  • Abliterated: the refusal direction is removed from the weights.
  • Uncensored: tuned without the filtering a general assistant ships with.
  • Fiction-tuned: trained for long-form writing and roleplay.

Each model has its own page with its price, and the proof page shows the daily check every model gets (310 of 398 passed on 2026-11-20). The whole price list is machine-readable at /v1/catalog.

Other tools

The same base URL and key work in OpenCode, Roo Code, Kilo Code, aider, Continue, Zed, LibreChat, Open WebUI, LiteLLM, n8n and SillyTavern. Each has a page at uncensoredllm.ai/guides.

Paying

Credit is prepaid from $10 by card, WeChat Pay, Alipay or crypto, and billed per token. There is no subscription required.

📰 Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.