Free forever · no credit card · no subscription

Put an AI assistant on any website. It answers from your content.

RagChat is a free, embeddable chat widget trained on your documents. Upload PDFs, Markdown, or text — connect your own AI key — and paste one line of code. Your visitors get instant, source-backed answers.

Platform cost
$0
Setup time
~5 min
Your data
Yours

Chat with Assistant

Online · answers from your docs

Hi there — ask me anything about this project.
What do you offer? How do I get started?
What do you offer?
I'm trained on your documents — upload PDFs, Markdown, or text and I'll answer from them. Completely free.
Type a message…
Send

Live preview — the widget your visitors will see

Free forever — the platform never bills you

Bring your own AI key — 16 providers supported

Answers come only from your own content

One script tag — works on any site

Features

Everything you need to ship an AI assistant

Every feature below is included and free. Here's precisely what RagChat can do — one feature at a time.

1 · Knowledge base — RAG over your documents

Core

Upload PDF, Markdown, or plain-text files — or paste content directly. RagChat splits every document into smart chunks (with context overlap so nothing gets lost at boundaries), converts them into embeddings, and stores them in a vector store. When a visitor asks a question, only the most relevant passages are retrieved and used to craft the answer. Your assistant answers exclusively from your content — never from general knowledge or thin air.

  • Upload .pdf .md .txt or paste text
  • Smart chunking with context overlap
  • Semantic retrieval — the top relevant passages win
  • Every answer grounded in a retrievable source

2 · Bring your own AI keys

$0

Connect your own API key from any of the 16 providers laravel/ai supports — OpenAI, Anthropic, Gemini, OpenRouter, DeepSeek, Mistral, Groq, Ollama, and more — then pick a chat model and a separate embedding model per project. RagChat never bills you: no subscriptions, no usage meters, no hidden fees. You pay your AI provider directly for the tokens you actually use, at their rates.

OpenRouter OpenAI Anthropic Gemini DeepSeek Mistral Groq Ollama Cohere xAI +6 more

The whole business model

The platform is 100% free. Your only cost is whatever your AI provider charges for the messages you send.

3 · One-line embed

Copy a single script tag and paste it before your closing </body> tag. No framework, no build step, no package install — it works on any static site, portfolio, or single-page app.

<script src="https://your-app.com/vendor/rag-chat/rag-chat.js" data-key="pk_…" data-mode="cdn" defer></script>

4 · Fully customizable widget

Rename the assistant, set a welcome message and input placeholder, choose from six hand-crafted themes, override the accent color, place the launcher left or right, and add up to six suggested messages visitors can tap to send — with a live preview as you edit.

5 · Two embed modes

CDN mode suits static sites — a public key protected by a domain allowlist. Backend mode keeps your signing secret on your server so the browser never sees it. Pick the one that fits your architecture.

6 · Streaming answers

Replies stream token-by-token, so visitors watch the assistant respond in real time — no long, silent waits while a full answer is generated.

7 · Domain verification & access control

Register your website, prove ownership with a one-line meta tag, and only verified domains can talk to your assistant. Remove a domain any time to revoke access instantly.

8 · Abuse protection & quotas

Set a daily message limit, benefit from per-IP cooldowns and layered rate limits, and issue short-lived signed sessions. Regenerate keys instantly if anything looks off.

9 · Full chat logs & analytics

Every embed message is logged with the visitor's IP, device fingerprint, embed mode, and timestamp — with total, daily, and weekly stats on your dashboard. Search and filter logs, then export them as CSV. (Chat logs auto-delete after 7 days.)

10 · Device fingerprinting & blocking

Every visitor gets a stable machine signature, so a banned user stays banned even after clearing cookies or switching networks. Block a device from your logs with one click — and unblock it just as easily if it was a mistake.

11 · Forgot password? Recover with a QR code

Enable two-factor authentication once — scan the QR code with any authenticator app (Google Authenticator, Authy, 1Password…). If you ever forget your password, enter your email and the 6-digit code to reset it instantly, with no waiting on support.

Themes

Six looks that match your brand

Every theme is a hand-crafted color system — plus a custom accent override for a perfect match with your site.

Swiss Light

Clean white panel, slate type, blue accent

/themes/swiss

Midnight

Dark charcoal with teal accent

/themes/midnight

Ocean Mist

Soft blue-gray atmosphere, deep ocean CTA

/themes/ocean

Forest

Sage surfaces with deep green accent

/themes/forest

Ink & Gold

High-contrast black with gold accent

/themes/ink

Stone

Neutral warm-gray studio look

/themes/stone

How it works

From zero to live assistant in three steps

No code required until the very last step — and even that is copy-paste.

1

Upload your content

Add PDFs, Markdown, or text files to your knowledge base — or paste content directly. RagChat chunks and embeds everything automatically, so it's instantly searchable.

2

Connect your AI key

Add a key from any of the 16 supported AI providers and choose your chat and embedding models. Your key, your rates — RagChat stays free.

3

Embed the widget

Verify your domain, copy the snippet, and paste it into any page. Your visitors instantly get a branded assistant that answers from your content.

Security & abuse control

Public on the web, protected by default

Your widget is public by design — but RagChat layers real protections on top so strangers can't drain your AI budget.

  • Domain allowlist — only meta-tag-verified sites can load your assistant.

  • Signed sessions — short-lived, HMAC-signed tokens bound to a project and origin.

  • Daily quotas — cap messages per day, plus per-IP cooldowns between messages.

  • Layered rate limits — per key, per IP, and per hour, tuned to stop scripted abuse.

  • Secret stays server-side — backend mode never exposes your signing key to the browser.

  • Instant key rotation — regenerate public keys and secrets with one click; old sessions die immediately.

  • QR-code recovery — two-factor authentication means you reset a forgotten password yourself, in seconds.

  • Machine-level blocking — banned devices stay banned across cookies, IPs, and browsers.

Pricing

Free forever. The platform costs $0.

RagChat doesn't sell plans, seats, or usage meters. Connect your own AI key and pay your provider directly for what you use — that's the entire business model. Every feature above is included, on the free tier, which is also the only tier.

No credit card · No subscription · Nothing to cancel

FAQ

Questions, answered

Is RagChat really free?
Yes. The platform is 100% free with no subscriptions, seats, or usage meters — and there is no paid plan. You connect your own AI provider API key and pay that provider directly for the tokens you use, at their published rates.
Which AI providers can I use?
All 16 providers supported by laravel/ai: OpenAI, Anthropic, Gemini, OpenRouter, DeepSeek, Mistral, Groq, Ollama, Cohere, xAI, Azure, Amazon Bedrock, and more. For each project you choose a chat model and a separate embedding model, so you can pair, say, a cheap chat model with a high-quality embedding model.
What file types can I upload?
PDF, Markdown, and plain-text files (up to 5 MB each) — or you can paste text directly. Every document is automatically chunked and embedded, so it is searchable the moment it is ingested.
Where does my data go?
Your documents and their embeddings are stored in your own RagChat project, and your assistant answers exclusively from that content. The AI provider you connect only sees the question, the retrieved passages you asked it to use, and generates the reply.
Can I use one assistant on multiple websites?
Yes. Register each website as a domain, verify it with a one-line meta tag, and the same knowledge base powers an assistant on all of them. You can remove a domain at any time to revoke access.
How do you prevent abuse on a public widget?
A combination of layers: a domain allowlist with meta-tag verification, a configurable daily message limit, per-IP cooldowns, layered rate limits per key/IP/hour, and short-lived signed session tokens. In backend mode, your signing secret never reaches the browser.
Do I need to know how to code?
Only to copy and paste. Uploading documents and connecting your key are point-and-click, and embedding is a single script tag. Backend mode includes copy-paste examples for developers who want maximum security.
What happens if I forget my password?
We never send password reset emails — instead, enable two-factor authentication once from your profile (scan the QR code with Google Authenticator, Authy, 1Password, or any other authenticator app). Then, if you ever forget your password, enter your email and the 6-digit code from your app to reset it instantly. No 2FA? Use the support route — we verify your identity and reset it for you, typically within 24 hours.
Can I see who is using my assistant?
Yes. Every message is logged with the visitor IP address, device fingerprint, embed mode, and timestamp. The dashboard shows all-time, daily, and weekly message counts, and the logs page lets you search, filter, and export everything as CSV. Chat logs auto-delete after 7 days, so export anything you want to keep.
Can I ban a specific visitor?
Yes — with a device fingerprint, not just an IP. Every visitor gets a stable machine signature, so blocking from your logs cuts that exact device off even if they clear cookies, change networks, or try another browser. Unblocking is one click.
Feedback

Help shape RagChat

Found a bug, want a feature, or just want to say hi? Every single submission is read by a human and feeds directly into what we build next.

  • Feature requests go on the roadmap

    The most-requested ideas get built first — your vote counts.

  • Bugs get fixed

    Reports are triaged and patched, then verified by our test suite.

  • We reply

    Leave your email and expect a real answer, not an autoresponder.

Send your feedback

Takes less than a minute.