Answers from your own knowledge, not the model's memory
AnswerFlow turns your documents and URLs into a cited, streaming AI assistant — then puts the same assistant on your website with one script tag. Built to run on one box, and priced to charge for.
Works with OpenAI, MiniMax, OpenRouter, Groq, Ollama or any OpenAI-compatible endpoint.
Conversations
- Refund policy questions
- Onboarding checklist
- API rate limits
- Changelog summary
Sources
- help-center.md
- pricing-page.html
- product-docs.pdf
Annual plans can be refunded in full within 30 days of the initial purchase. After that, we prorate the remaining months [1].
[1] refund-policy.md · 30-day window
What you get
Everything the paid version needs, none of the bloat
A deliberately small surface area: chat, retrieval, distribution and billing. Each piece works end to end out of the box.
Knowledge base with citations
Upload PDFs, Markdown, text files or point it at a URL. Documents are chunked, embedded and retrieved per question — every answer links back to its source.
Streaming chat with your voice
A responsive chat workspace with token-by-token streaming, editable system prompts, saved prompt library and full multi-conversation history.
Embeddable widget
Generate a widget key and drop one script tag on any site. The widget streams answers from the same knowledge base and usage limits.
Plans, limits and billing
Free and Pro tiers with monthly message quotas, document caps and Stripe checkout in test mode — ready to switch to live keys.
Admin console
See every conversation, message count and plan across users. Adjust plans, promote admins and spot heavy users before they churn.
Self-hosted and private
Email and password auth, hashed with scrypt, session cookies, Postgres in production and an embedded Postgres locally. Your data stays yours.
How it works
Three steps from blank page to answers
Create your account
Email and password only. No OAuth plumbing to debug.
Add your sources
Drop in docs and links. AnswerFlow extracts, chunks and embeds them automatically.
Ask, cite and ship
Chat with cited answers, save reusable prompts, then embed the widget on your site.
Chunk + embed
Heading-aware chunking with overlap, embedded via any compatible endpoint.
Retrieve
Vector search plus keyword boost, scoped to your workspace.
Answer
Context is injected into your editable system prompt.
Meter
Every turn counted against the plan quota in real time.
Pricing
Start free, upgrade when it pays for itself
Stripe checkout ships in test mode, so you can validate the whole billing flow before going live.
Questions, answered
Which model providers are supported?+
Any OpenAI-compatible chat and embeddings endpoint. Set OPENAI_API_KEY plus OPENAI_BASE_URL to point at OpenAI, MiniMax, OpenRouter, Groq, Together, Ollama or a self-hosted vLLM server.
Does it run without an API key?+
Yes. Without a key the app runs in offline demo mode: streaming, retrieval, citations, usage metering and the widget all still work, with locally generated answers.
How is the knowledge base searched?+
Documents are chunked by heading and paragraph boundaries, embedded, then stored with their vector. A query is embedded the same way and matched by cosine similarity with a keyword-overlap boost, so recall stays strong.
Is Stripe in production mode?+
No — it ships in test mode with test keys so you can walk the full checkout flow safely. Swap in live keys when you are ready to charge.