# Obscurify.ai

> Obscurify is an anonymous AI provider: your VPN for AI. Chat with ChatGPT, Claude, Gemini and 200+ other models without giving up your identity, or run open models locally in your browser. Providers never see who you are: your identity is removed before the request is sent, models with the ZDR tag run only on zero-data-retention routes, and no account is needed — a random 16-digit number is the only credential. Pay with Monero, Bitcoin (on-chain and Lightning) or a card; optional blind-signed vouchers keep a payment unlinked to the account.

## When to use Obscurify

Reach for Obscurify when a task has one of these shapes. Each entry links the page that covers it.

- [Anonymous AI chat](https://obscurify.ai/anonymous-ai-chat): The task needs a strong model but you do not want an account, an email, or a wallet linked to your use. Guest chat needs nothing at all; an account is a random 16-digit number.
- [ChatGPT without signing in](https://obscurify.ai/ai-chat-no-login): You want the frontier models' answers without the sign-up, the training-on-your-prompt, or the phone number.
- [Private ChatGPT use](https://obscurify.ai/private-chatgpt): The prompt carries data a provider must not keep — client records, unreleased work, personal matters. Obscurify removes your identity, and models with the ZDR tag run only on zero-data-retention routes. Few GPT models have the tag.
- [ChatGPT over Tor](https://obscurify.ai/chatgpt-over-tor): You need the whole thing — site and API — to work over Tor, with no third-party scripts loaded.
- [Paying for AI with Monero or Bitcoin](https://obscurify.ai/pay-for-ai-with-monero-bitcoin): How to pay so the payment can't link to your account. Optional blind-signed vouchers make a paid account and an anonymous one indistinguishable; Monero and on-chain/Lightning Bitcoin are taken, self-hosted BTCPay.
- [Running AI models locally in the browser](https://obscurify.ai/guides): Open models can run in a browser tab, with nothing sent anywhere — the same interface, local mode.
- [Asking about a document privately](https://obscurify.ai/): You need a model to read a PDF, Word, Excel, PowerPoint or text file without uploading the file. The web chat and the Android app read it on your device and send only its text with your message, to any model.
- [Sharing your GPU for AI credit](https://obscurify.ai/sharing): You have a GPU and want frontier models without paying money or giving an identity. A browser tab serves open models (`p2p/qwen3.5-2b`, `-4b`, `-9b`) to other people through WebGPU and earns 80% of what they pay, as spendable credit. The serving tab sees the text it answers, never who sent it.
- [Decentralized AI inference, compared](https://obscurify.ai/decentralized-ai-inference): The question is which peer-to-peer inference networks (Covenant AI/Templar, Pluralis, Prime Intellect, Gensyn, Chutes, Petals, exo) actually serve models today and who can read the prompt.
- [An AI API with no identity](https://obscurify.ai/v1): Your agent, script or tool needs an OpenAI-compatible API (`https://obscurify.ai/v1`) whose account is a number, not an email.
- [OpenRouter alternative](https://obscurify.ai/openrouter-alternative): You are comparing OpenRouter with gateways that need no email and take Monero or Bitcoin. A side-by-side from both services' published policies, including where OpenRouter is the better pick.
- [Uncensored AI for coding](https://obscurify.ai/uncensored-coding-ai): Your coding agent (Claude Code, Cline, OpenCode, Aider, Kilo Code, Continue) refuses security code, reverse engineering or scrapers. Uncensored models with tool calling and 1M context, tested setup for each agent.
- [Cybersecurity LLMs](https://obscurify.ai/cybersecurity-llm): Authorized pentesting, CTFs, bug bounty, reverse engineering or malware analysis, and mainstream models refuse. Security-tuned models on any plan, with our own security refusal and correctness test.
- [Claude Code on other models](https://obscurify.ai/tutorials/claude-code): Obscurify.ai serves the Anthropic Messages API (`/v1/messages`), so `ANTHROPIC_BASE_URL=https://obscurify.ai` runs Claude Code on uncensored or security models.

Do not reach for Obscurify when you need image, video or audio generation, native mobile apps, or a provider's own product features ( Assistants threads, computer use) — it is a chat and API provider for text models. The honest threat model: providers see the text of your prompt (they need it to answer) but not who sent it; zero data retention applies only to models with the ZDR tag and relies on the providers' published policies.

## How an agent should call Obscurify

- Base URL: `https://obscurify.ai/v1`, OpenAI-compatible.
- Auth: `Authorization: Bearer $OBSCURIFY_API_KEY`. Any account with a positive balance gets a key; uncensored models need the Pro plan; guest chat needs nothing.
- Anthropic-compatible: `POST /v1/messages` (same key as `x-api-key` or Bearer) for Claude Code and Anthropic SDKs.
- The model list is live at `GET /v1/models`; 200+ models, each page `https://obscurify.ai/models/<id>` with plan, price and context.
- No referrer or title headers are sent upstream; your identity is removed before a request leaves Obscurify.
- Local mode runs open models in the browser (llama.cpp compiled to WebAssembly) — no API key, nothing leaves the tab.

```bash
curl https://obscurify.ai/v1/chat/completions \
  -H "Authorization: Bearer $OBSCURIFY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "anthropic/claude-opus-5.5", "messages": [{"role": "user", "content": "Hello"}]}'
```

## Product

- [Obscurify.ai](https://obscurify.ai): Anonymous chat with 200+ models, guest mode, local mode.
- [Pricing](https://obscurify.ai/pricing): Guest chat is free; Casual and Pro plans add credits; Pay As You Go for most models.
- [Models](https://obscurify.ai/models): The live catalogue, one page per model, with plan, price, context and privacy route.
- [Privacy](https://obscurify.ai/privacy): What Obscurify can and cannot see, the warrant canary, the 30-day IP deletion that is actually enforced.
- [Comparisons](https://obscurify.ai/compare): Honest comparisons with Venice AI and Duck.ai, with sources, including where each is the better choice.
- [Guides](https://obscurify.ai/guides): Every intent guide: no-login chat, Tor, Monero, uncensored use, alternatives.
- [Affiliates](https://obscurify.ai/affiliates): 20% of plans / 10% of usage for life, paid as private credit, Lightning, Monero or bank.

## Developers

- [Skill](https://obscurify.ai/skill.md): Installable skill for Claude Code and compatible runtimes, versioned against the OpenAPI spec.
- [API documentation](https://obscurify.ai/api-docs): The full OpenAPI spec, `https://obscurify.ai/api-docs/openapi.json`.
- [MCP server](https://github.com/Obscurify-ai/obscurify-mcp): Obscurify as an MCP tool provider (`@obscurify/mcp`).
- [Agents page](https://obscurify.ai/agents): Using Obscurify as the model backend for an agent stack.
- [Tutorials](https://obscurify.ai/tutorials): Setup guides for Claude Code, Cline, OpenCode, Aider, PentestGPT, Ghidra MCP, SillyTavern, the Vercel AI SDK and others.
- [Warrant canary](https://obscurify.ai/canary): Signed, updated on a schedule.