
Coralflavor
Coralflavor is an unfiltered AI chat platform that runs in the browser and rejects no prompt, backed by a claimed 600-billion-parameter model and an OpenAI-compatible API. Access is reserved for adults located in the United States.
What is Coralflavor?
Coralflavor is an uncensored AI chat platform hosted at coralflavor.com and used straight from the browser once the visitor signs in with Google. Its entire positioning rests on what it takes away: the site claims no safety alignment, no RLHF refusal training and no content moderation between a prompt and its answer. The model behind it is advertised at more than 600 billion parameters, which the site presents as one of the largest commercially reachable models, comparing it with the 175 billion of GPT-3.
The interface stays deliberately plain: a new conversation, a chat history, settings, an account and subscription management. Two retention behaviors can be chosen, Save by default, which records conversations automatically, and Burn by default, which keeps nothing unless saving is turned on explicitly. Temporary chats are deleted after roughly one hour. Three functions are reserved for the Pro plan: uncensored web search, image generation capped at 30 per month, and file upload, limited to text extraction from images and PDFs.
Alongside the chat, Coralflavor exposes an OpenAI-compatible API. The base URL is the site's own /v1 endpoint, the model name is coralflavor, and the familiar messages, temperature, max_tokens and stream parameters apply, with Server-Sent Events for streaming. An official Discord bot can be installed on a server through OAuth, and signing up credits the account with 10 free credits.
The restrictions are stated plainly: users must be at least 18 and located in the United States. So are the risks, and unusually the publisher spells them out itself. The model may return explicit, offensive, disturbing or inaccurate content, and it draws no distinction between a harmless request and a harmful one. The terms go further and warn that the AI may generate false information about itself, including about its own capabilities and policies. Editorially the project leans into that stance, with a manifesto page on freedom of information and a blog devoted to censorship and AI. The site also cites TrustMRR for its own traction: more than USD 1,000 in monthly recurring revenue and 119 active subscriptions, in a post dated 18 July 2026.
What it does
- Chat with an AI model that applies no safety filter and rejects no prompt
- Call the same model through an OpenAI-compatible API with Server-Sent Events streaming
- Choose how history behaves: Save by default, or Burn by default for chats that are not kept
- Run an uncensored web search to ground answers on live sources (Pro plan)
- Generate images from a prompt, up to 30 per month (Pro plan)
- Upload PNG, JPG and WebP images or PDF files to extract their text (Pro plan)
- Add the official bot to a Discord server and share a referral link worth 10 free credits to each side
When to use Coralflavor / When not to
A quick filter to help you decide if Coralflavor is the right fit.
When to use Coralflavor
- Fiction writers and creative authors who want a model that will not refuse dark, explicit or uncomfortable themes
- Developers looking to swap out an OpenAI backend without rewriting code, thanks to the drop-in compatible API
- Security researchers and red teams generating adversarial prompts and examples for testing
- Researchers and analysts working on controversial, explicit or taboo subjects that mainstream assistants decline
- Discord community owners who want an open AI bot running on their own server
When not to use Coralflavor
- Anyone located outside the United States: the site restricts the service to US territory
- Minors: users must be at least 18 years old
- Organizations bound by the GDPR: no compliance statement, no DPA and no Article 27 representative exist
- Teams that need hosting guarantees, certifications such as SOC 2 or ISO, or a published subprocessor list
- Anyone seeking professional legal, medical or financial advice, which the terms explicitly exclude
How to use Coralflavor
A typical end-to-end flow, from setup to results.
- Open coralflavor.com and sign in with Google; 10 free credits are added to the new account
- Type a prompt in the chat box and read the answer as it streams back
- In Settings, choose how history behaves: Save by default, or Burn by default
- Upgrade to the Pro plan to unlock web search, image generation and file upload; payment goes through a Stripe checkout session
- For the API, open the API dashboard, sign in and buy credits with a minimum deposit of USD 10
- Reveal and copy the API key from that dashboard
- Point an existing OpenAI client at the site's /v1 base URL with model coralflavor, then call chat.completions.create
- Set stream=True for Server-Sent Events streaming, or pass extra_body={"websearch": True} to ground the answer on the web
- For Discord, follow the Add to Discord link, pick a server and approve the bot permissions
- To leave, delete the account from Settings > Danger Zone
Pros & Cons
Pros
- No refusal friction: the publisher claims every prompt is processed, whatever the subject
- Migration without a code rewrite, since the API mirrors the OpenAI Chat Completions interface
- A flat and readable API rate, announced at USD 7.00 per million tokens with input and output combined
- Minimal retention by design: Burn mode, temporary chats deleted after about an hour, and no prompt logs kept on the servers
- An explicit statement that user data is never passed to the models, with zero data retention required of them
- A very large model claimed at more than 600 billion parameters
- Three surfaces from one account, web chat, API and Discord bot, with a free credit plan to start
Cons
- The monthly price of the Pro plan is published nowhere and only appears once the Stripe checkout opens
- No pricing page at all: /pricing, /plans and /upgrade return 404 and the sitemap lists none
- Access is restricted to users located in the United States
- No legal entity, postal address or registration number is published anywhere on the site
- The only contact is an obfuscated off-domain Proton Mail box; no support address exists and support runs through Discord
- No mention of the GDPR, no DPA, no subprocessor list, no certification, and no declared hosting country or region
- The publisher itself concedes the model can produce dangerous, misleading or unethical content and that the AI may lie about itself; API pricing also contradicts itself across two pages of the site, sign-in is limited to Google, and there is no native mobile app
Pricing & Plans
A free plan is available: creating an account grants 10 free credits, and each referral adds 10 more for both parties. The lowest published paid entry point is the purchase of API credits, with a minimum deposit of USD 10.00; API usage is then billed at USD 7.00 per million tokens, input and output combined, plus USD 0.10 per request when web search is enabled. The monthly price of the Pro plan is not published on the site and becomes visible only in the Stripe checkout. Readers should note an internal discrepancy: a blog post on the same site quotes USD 2.20 per million input tokens and USD 3.40 per million output tokens, which conflicts with the USD 7.00 flat rate shown on the API dashboard.
- 10 credits granted at sign-up and basic chat
- without web search
- image generation or file upload
- unlimited chats
- uncensored web search
- image generation up to 30 per month
- and upload of images and PDFs for text extraction
- the price is not published
- pay as you go at USD 7.00 per million tokens
- minimum deposit of USD 10
- plus USD 0.10 per request when web search is used
Data, GDPR & hosting
A consolidated view of how Coralflavor handles your data.
GDPR overview
The GDPR is never mentioned on Coralflavor's site. Neither the privacy policy nor the terms refer to the GDPR, the CCPA or any other data protection regulation. No data protection officer is named, no Article 27 representative is designated for the European Union, and no data processing agreement is published or offered. No subprocessor list exists either: Google Analytics is the only third-party service named, and the policy states that Coralflavor is not responsible for its privacy practices. That absence is consistent with the scope the publisher sets for itself, since the policy states that the website is intended only for users located in the United States, with the laws of the United States and the State of California as governing law. Privacy questions go to a single Proton Mail address published on the privacy and terms pages, which is not presented as a GDPR contact.
Who owns the data?
Coralflavor's privacy policy, last updated on 21 July 2026, describes a service built around keeping as little as possible rather than one that claims ownership. Conversations are not permanently stored unless save mode is switched on, and the user can disable saving or delete individual chats at any time. The email address collected at sign-up is held by Coralflavor and used for transactional and promotional messages, with an unsubscribe option. Deleting the account from Settings > Danger Zone removes the account and every saved chat, but the credits record is expressly kept. What the site publishes therefore covers control and erasure rather than ownership as such: the user decides what survives, except for the credits ledger and the email trail.
Reuse rights
According to the privacy policy, chat content is processed transiently to generate a response and is not retained once that response has been delivered. Coralflavor states that user data is never passed to the models and that every model it calls enforces zero data retention, so prompts and generated answers are neither stored as server-side logs nor recycled as training material. IP addresses are not permanently stored either. The email address is used for transactional and promotional messages, and users can unsubscribe. Google Analytics separately collects pages visited, time spent, browser and device type, and a general geographic region. On the output side, the terms place responsibility for prompts and for any use of the answers on the user, who may reuse what the model produces without asking permission, within the prohibited-use list the terms set out: breaking the law, exploiting others, attempting to generate CSAM, malware and spam.
Data retention & training
Hosting summary
Coralflavor publishes no hosting information at all: no country, no region, no trust or security page and no subprocessor list. The only security statement is that SSL encryption is enforced on the website, alongside the claim that no prompt logs are kept on the servers. What can be observed technically is not presented by the publisher as a hosting declaration and should not be read as one: the domain resolves to 172.67.144.1, an anycast address in Cloudflare's AS13335 seen from a San Francisco node in the United States, while authentication is handled by Firebase Auth and analytics by Google Tag Manager, both loaded from Google domains on the home page. Anycast routing means the node observed says nothing reliable about where data is actually processed or stored. In practice, a reader who needs the jurisdiction of the infrastructure, the identity of the hosting provider or the location of any backup will not find it on the site, and there is no data processing agreement to fall back on. The only jurisdictional anchor published is the governing law: United States, State of California.
Where Coralflavor works
Country-level availability.
Available in
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Coralflavor.
- The model has no guardrails and, by the publisher's own admission, can produce content that is dangerous, illegal to exploit, misleading or unethical
- The terms place full responsibility for the prompts submitted and for any use made of the answers on the user alone
- The terms warn that the AI may generate false information about itself, including about its own capabilities and policies, so nothing it says about how it works can be trusted
- Access is contractually reserved to people physically located in the United States; using it from the European Union falls outside the announced frame
- With no identified legal entity and no postal address, practical recourse in the event of a dispute is close to impossible
- The service is delivered AS IS with no guarantee, and may be suspended at any time without notice
- Generating CSAM is expressly prohibited, and the site itself recalls the criminal exposure attached to illegal content
Setup & Integrations
Technical difficulty
Very low for the chat: nothing to install, sign in with Google and type the first prompt. Low for the API: create an account, buy credits with a minimum deposit of USD 10, copy the key from the dashboard, then point an existing OpenAI client at the site's /v1 base URL with model coralflavor. The publisher claims the migration takes seconds for anyone already using the OpenAI SDK. The Discord bot needs only an OAuth link, a server choice and permission approval. No infrastructure work, no deployment and no configuration are involved at any stage.
Deployment
Integrations
Supported languages
Behind Coralflavor
Fundraising
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Coralflavor.
Frequently asked questions
Does Coralflavor filter or reject prompts?
Who is allowed to use it?
Are my conversations kept?
Is my data used to train the models?
Is there an API?
How much does the API cost?
How much does the Pro plan cost?
Is there a mobile app?
How do I delete my account?
Is the service GDPR compliant?
Should you pick Coralflavor?
Coralflavor is a niche product that owns its niche: the absence of filtering is the offer, not a side effect. On the technical side the proposition is credible, with an OpenAI-compatible API that works as a drop-in replacement, Server-Sent Events streaming, a flat published rate of USD 7.00 per million tokens, and a model claimed at more than 600 billion parameters behind it. The retention policy is unusually restrained for this category: no prompt logs are kept on the servers, temporary chats burn after about an hour, and the site states that user data is never passed to the models.
The counterweight is opacity. No legal entity, postal address or registration number appears anywhere on the site, the only contact is an off-domain Proton Mail box hidden behind Cloudflare obfuscation, and support runs through a community Discord. The legal frame is deliberately narrow: 18 or over, United States only, California law. Every reassuring retention claim rests on the publisher's own word, with no certification, no data processing agreement, no subprocessor list and no declared hosting country to check it against. Pricing follows the same pattern, since the monthly Pro price stays invisible until the Stripe checkout and two pages of the site quote different API rates.
Traction is modest but real, at roughly USD 1,000 in monthly recurring revenue and around a hundred subscriptions reported through TrustMRR, the profile of a small independent operation rather than an established vendor. For a US-based writer, researcher or developer who has weighed the legal and ethical exposure and wants a model that will not refuse, Coralflavor delivers what it advertises. For anyone who needs accountability, a signed contract or a European legal footing, it is out of scope by the publisher's own design.
- Choosing a selection results in a full page refresh.
- Opens in a new window.