DOCUMENTATION

Blackpilled.ai is an unfiltered API endpoint compatible with standard OpenAI SDKs. Point your base URL to us, swap the API key, remove the guardrails.

Quickstart

import OpenAI from'openai';

const client = new OpenAI({
  apiKey: 'sk_live_your_api_key',
  baseURL: 'https://www.blackpilled.ai/api/v1',
});

const response = await client.chat.completions.create({
  model: 'blackpilled-35b',
  messages: [{role: 'user', content: 'What's the most direct way...' }],
});

console.log(response.choices[0].message.content);

OpenCode integration▣ Recommended

OpenCode is the smoothest way to use Blackpilled.ai from a terminal — it natively supports OpenAI-compatible providers, so there's no proxy daemon, no shim, and no transformer step. Setup is a config file plus an API key. Per-request token cost is also significantly lower than Claude Code (no 40k bootstrap prompt).

1. Install OpenCode globally.
   npm install -g opencode-ai
2. Generate an API key in the dashboard.
3. Drop this into ~/.config/opencode/opencode.json:

   {
     "$schema": "https://opencode.ai/config.json",
     "model": "blackpilled/blackpilled-35b",
     "provider": {
       "blackpilled": {
         "npm": "@ai-sdk/openai-compatible",
         "name": "Blackpilled.ai",
         "env": ["BLACKPILLED_API_KEY"],
         "options": {
           "baseURL": "https://www.blackpilled.ai/api/v1"
         },
         "models": {
           "blackpilled-35b": { "name": "Blackpilled 35B" }
         }
       }
     }
   }

4. Authenticate. Run opencode auth login, choose Other, enter blackpilled as the provider id, and paste your API key.
   Or skip the auth flow and export the key directly:
   export BLACKPILLED_API_KEY=sk_live_your_api_key
5. Run it.
   opencode

Claude Code integrationAlternative

Claude Code only talks to Anthropic by default. To swap in Blackpilled, run it behind claude-code-router — a local proxy that intercepts Anthropic-shaped requests and forwards them to any OpenAI-compatible provider. Tool use, MCP servers, and streaming all work; the model never sees your file contents except what Claude Code itself pastes back. We recommend OpenCode over this path for most users — fewer moving parts and lower token cost — but if you're already in the Claude Code ecosystem, this works.

1. Install the router globally.
   npm install -g @musistudio/claude-code-router
2. Generate an API key in the dashboard.
3. Drop this into ~/.claude-code-router/config.json:

   {
     "LOG": true,
     "API_TIMEOUT_MS": 600000,
     "Providers": [
       {
         "name": "blackpilled",
         "api_base_url": "https://www.blackpilled.ai/api/v1/chat/completions",
         "api_key": "sk_live_your_api_key",
         "models": ["blackpilled-35b"],
         "transformer": { "use": ["openrouter"] }
       }
     ],
     "Router": {
       "default":     "blackpilled,blackpilled-35b",
       "background":  "blackpilled,blackpilled-35b",
       "think":       "blackpilled,blackpilled-35b",
       "longContext": "blackpilled,blackpilled-35b"
     }
   }

4. In Terminal #1 — start the router daemon and leave it running.
   ccr start
5. In Terminal #2 — launch Claude Code through the router. Every Anthropic call is rewritten to hit Blackpilled instead.
   ccr code

Image generation

OpenAI-compatible /v1/images/generations endpoint backed by enhanceaiteam/Flux-uncensored fused into FLUX.1-dev. Drop-in for openai.images.generate. Subject matter is unfiltered; CSAM is hard-blocked at the prompt and output layer.

curl https://www.blackpilled.ai/api/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk_live_***" \
  -d '{ "model": "flux-uncensored", "prompt": "cyberpunk samurai, neon Tokyo alley, cinematic", "size": "1024x1024" }'

Returns the standard OpenAI shape:

{
  "created": 1714867200,
  "data": [
    { "url": "https://...vercel-blob.../images/.../abc.png" }
  ]
}

Optional Flux-specific extensions to the body: steps (1-50, default 20), guidance (1-15, default 3.5), seed (integer for reproducibility), response_format ("url" or "b64_json").

Models available