API docs

unfiltered.sh speaks the OpenAI Chat Completions API. If a tool lets you set a base URL and an API key, it works with unfiltered.sh.

Quick start

1. Create an account, add credit with SOL or USDC, and buy a seat (a slot plus a speed).
2. Copy the API key shown after purchase. It is shown once.
3. Call the API during your slot with these settings:

Models

Use the model id in the model field of each request. GET /v1/models lists what your seat can use.

DeepSeek V4.1 Flash is a community "abliterated" build of an open-weight mixture-of-experts model (refusal behaviour removed). We did not train it.

Examples

curl https://api.unfiltered.sh/v1/chat/completions \
  -H "Authorization: Bearer $UNFILTERED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash",
    "messages": [{"role": "user", "content": "Explain how a TLS handshake works."}],
    "stream": true
  }'

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.unfiltered.sh/v1", api_key=os.environ["UNFILTERED_API_KEY"])

stream = client.chat.completions.create(
    model="deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Write a haiku about GPUs."}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="", flush=True)

const res = await fetch("https://api.unfiltered.sh/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.UNFILTERED_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "deepseek-v4.1-flash",
    messages: [{ role: "user", content: "Summarise the CAP theorem in 3 lines." }],
  }),
});
const data = await res.json();
console.log(data.choices[0].message.content);

Limits

• Slot window. A key works only between its slot's start and end time. Before the start you get block_not_started, after the end block_ended. Requests still streaming when the slot ends may be cut off.
• One request at a time per seat. Your speed is reserved for one stream. A second parallel request gets HTTP 429 concurrency_limit; queue requests on your side or buy a second seat.
• Speed. Your chosen tok/s is a per-request target for generated tokens. Responses carry an x-seat-tps header; if the cluster is overloaded you may see x-speed-degraded: true, and missed speed is credited back.
• Unlimited tokens, with fair-use guards. You are not billed per token. To protect other seats there is a per-minute request limit, a maximum output length per request, and a very high per-seat output safety cap that normal use does not reach.
• No logs. Prompts and outputs are never stored. Only token counts and timings are kept for billing and speed stats.
• Acceptable use. Requests that break the Acceptable Use Policy are blocked with policy_violation.

Errors

Errors use the OpenAI shape: {"error": {"message", "type", "code"}}.

Use it with coding agents

Any agent that supports an "OpenAI-compatible" provider works. Use the base URL, your seat key and the model id above. Because a seat runs one request at a time, turn off parallel sub-agents or set concurrency to 1.

Aider

export OPENAI_API_BASE=https://api.unfiltered.sh/v1
export OPENAI_API_KEY=YOUR_API_KEY
aider --model openai/deepseek-v4.1-flash

Cline, Roo Code, Kilo Code (VS Code)

Settings → API Provider: OpenAI Compatible. Base URL https://api.unfiltered.sh/v1, API key = your seat key, Model ID deepseek-v4.1-flash.

Continue

models:
  - name: unfiltered.sh
    provider: openai
    model: deepseek-v4.1-flash
    apiBase: https://api.unfiltered.sh/v1
    apiKey: YOUR_API_KEY

Anything else

Set the environment variables OPENAI_BASE_URL=https://api.unfiltered.sh/v1 and OPENAI_API_KEY=YOUR_API_KEY; most tools built on the OpenAI SDKs pick them up.