Obscurify.ai

Core positioning

Obscurify.ai describes itself as an anonymous AI provider and a “VPN for AI.” Its positioning is Private, Local, and Uncensored. The service claims no logs, no tracking, no identity requirements, and no conversation storage, while keeping users' identities hidden from cloud model providers.

Main features

• Anonymous cloud chat: Access ChatGPT, Claude, Gemini, and other models through one interface. The product description advertises more than 200 models, while the model directory lists 299, including providers such as OpenAI, Anthropic, Google, Meta, DeepSeek, and Qwen.
• Free chat without login: Guests can use Meta's Llama 3.3 70B Instruct without an account, email address, or cookies. The directory lists one free model.
• Anonymous accounts: An account consists only of a random account number, with no email, username, or password required. Registration opens access to additional model options, while frontier and uncensored models remain subject to plan permissions.
• Local browser-based models: Run open-source GGUF models entirely on the user's device. Prompts never leave the device in local mode. The initial model download may take a few minutes.
• Tor access: The entire website works through a Tor hidden service, allowing users to access it without revealing their IP address to the site.
• Model discovery and favorites: Filter by provider or categories such as cheap, reasoning, uncensored, roleplay, and frontier models. Users can save up to 20 favorite models.
• OpenAI-compatible API: The API works on every plan for any account with a balance, supporting applications and development workflows compatible with the OpenAI API.
• Additional access points: The interface provides a Web Search tool entry and an Android app link.

Typical workflow

1. Open the website and chat with the free Llama 3.3 70B Instruct model as a guest, without registering.
2. For additional cloud models or API usage, create a random-number account, add a prepaid balance or choose a subscription, and use models according to their access requirements.
3. For local inference, open Local Mode, select and download a GGUF model, then use Download & Enable to run it on the device.
4. For Tor access, use the service's Onion site.

Pricing and payments

• Free access: Guests can chat with Llama 3.3 70B Instruct for free.
• Pay as you go: Usage can be charged against a prepaid balance. Individual model rates are not specified.
• Subscriptions: Plans start at $20 per month. The Pro plan adds uncensored models; a separate Pro price is not specified.
• Promotional API credit: New accounts may receive $1 in starter credit while the current month's promotional budget lasts. This is not guaranteed for every new account.
• Payment methods: Monero, Bitcoin, Lightning, and cards are supported, with payments of any amount advertised. The service says it uses blind vouchers so it cannot associate a particular payment with the account it funds.

Intended users and use cases

• Privacy-conscious users who want AI chat without providing an email address or login details.
• Users seeking a unified interface for multiple models, with selection by reasoning capability, roleplay category, or price.
• Users who want prompts to remain entirely on their device through local browser inference.
• Users who need to access AI services through Tor.
• Developers seeking an OpenAI-compatible API with prepaid billing and multiple model options.

Important distinctions

Cloud mode emphasizes identity shielding and no conversation storage, whereas local mode explicitly keeps prompts on the device. These are different data-handling approaches. The open-source description applies to supported GGUF models, not necessarily to the Obscurify.ai platform itself. Free guest chat, paid model access, and the API balance requirement have separate usage conditions.