Bring your own key · Beta
Run Sonny's AI on your own account
Connect your own OpenAI or Anthropic (Claude) key — or any OpenAI-compatible endpoint, including a self-hosted Ollama server. Unlimited AI conversations at raw token prices, for a one-time $249.
What you get
$249, once, forever
One purchase per workspace unlocks the whole thing. No new subscription, nothing held back.
- Unlimited AI conversations
- No monthly allowances, no overage anxiety. Your provider bills you for exactly the tokens you use — Sonny never meters your conversations again.
- Any provider, any model
- OpenAI, Anthropic (Claude), or any OpenAI-compatible endpoint — including a self-hosted Ollama or vLLM server. Pick a curated model or enter any model ID.
- AI processing where you choose
- Made for teams on custom server environments or with data-residency requirements. Every AI response is generated on your account, under your provider agreement.
- Encrypted, verified credentials
- Your API key — and any proxy credentials — are encrypted at rest (AES-256-GCM) and never shown again. Connections only go live after a real connectivity and tool-calling test passes.
- Every AI feature included
- The 24/7 auto-responder, AI reply drafts for your team, and one-click polish all run through your key. Nothing is held back.
- Pay once, keep it forever
- The unlock is a one-time $249 purchase per workspace — not another subscription. Switch between your key and Sonny AI plans whenever you like.
The math
It pays for itself fast
A typical AI support conversation costs about a cent in API tokens on an efficient model like Claude Haiku. Next to monthly plan pricing, the unlock usually earns its keep within the first month or two at higher volumes.
| Monthly volume | Sonny AI plan | With BYOK* |
|---|---|---|
| 300 conversations | AI Starter — $29.99/mo | ~$3/mo |
| 1,500 conversations | AI Growth — $79/mo | ~$15/mo |
| 5,000 conversations | AI Scale — $199/mo | ~$50/mo |
| 20,000 conversations | Not available on plans | ~$200/mo — no cap |
*Estimated API cost billed by your provider, based on Claude Haiku 4.5 pricing and an average support conversation (~3 AI replies). Actual costs depend on your model, knowledge-base size, and conversation length. Self-hosted models cost whatever your hardware costs.
Running 5,000 conversations a month? That's roughly $199/mo on a plan vs ~$50/mo on your own key — the unlock pays for itself in the second month, and everything after that is savings.
Still managed
You bring the key, we do the heavy lifting
BYOK swaps out who generates the AI's responses — nothing else changes. Your knowledge base, embeddings, and retrieval pipeline stay fully managed by Sonny, at no extra cost.
- Website crawling and re-crawling — we keep paying the crawler bills
- Document processing and embeddings — your knowledge base just works
- Hybrid retrieval (vector + keyword search) tuned for support conversations
- Missed-question capture, test chat, confidence thresholds, and human handoff
- Automatic fallback to your team if your provider ever has a bad day
You can feed the assistant your own business documents and website content (RAG) exactly as before — via the dashboard or the built-in chat widget — while every customer-facing response is generated on infrastructure you choose.
Providers
Works with what you already trust
- OpenAI
- GPT models on your OpenAI account. Curated picks for support work, or any model ID you prefer.
- Anthropic (Claude)
- The same model family Sonny AI runs on. Claude Haiku is our recommended default — fast, accurate, and inexpensive.
- Self-hosted & compatible
- Any OpenAI-compatible endpoint: Ollama, vLLM, LiteLLM, or your company's internal gateway. Your models, your servers, your rules — reachable over HTTPS, and protected by HTTP Basic Auth or your gateway's own access headers.
Setup
Live in five minutes
Four steps, all from the AI Copilot page in your workspace.
- 01
Unlock BYOK
A one-time $249 purchase from the AI Copilot page. Owners only, takes a minute.
- 02
Connect your provider
Paste your OpenAI or Anthropic key, or point Sonny at your own OpenAI-compatible endpoint over HTTPS. If a proxy guards that endpoint, add its credentials instead of an API key.
- 03
Test before it goes live
Sonny verifies your key can authenticate and handle tool calls — the connection can't be enabled until the test passes.
- 04
Flip the switch
All AI responses now run on your account. Toggle back to Sonny AI anytime — your setup is saved.
AI on your own terms
Unlock Bring your own key from the AI Copilot page. One-time $249 per workspace, yours forever.