BYOK AI Support Widget

AI customer support
for $29/mo,
not $1,019.

Use your own OpenAI or Anthropic key. Pay wholesale API rates — not Intercom's $0.99 per resolution markup.

A fully customizable chat widget that embeds natively into your site, calls your backend tools, and stops penalizing you for every conversation your AI resolves.

Early access

Join the waitlist

We'll email you when the beta opens. No spam, no sales calls.

You're on the list!

We'll be in touch when the beta opens.

Free to start — no credit card required

Intercom charges $0.99 every time your AI helps a customer.

That sounds small. It isn't.

At 1,000 conversations a month, you're paying $1,019 — $990 in "AI resolutions" plus a $29 seat fee. At 5,000 conversations: $5,029. At 10,000: $9,929.

The more your AI improves, the higher your bill. That's not a pricing model. It's a trap.

The actual cost to run that conversation through GPT-4o mini?

About half a cent.

$ intercom --conversations 1000 seat fee: $29.00 ai resolutions: $990.00 (1000 × $0.99) ────────── total: $1,019.00 $ supportkit --conversations 1000 platform fee: $29.00 api (gpt-4o-mini): $3.00 ────────── total: $32.00

Bring your own API key. Keep the savings.

Most AI support platforms are reselling OpenAI's API at a 200× markup and calling it "AI pricing." With SupportKit, you connect your own key and pay your provider directly — at wholesale rates. We charge a flat monthly fee for the platform. That's it.

Conversations / mo Intercom cost Your API cost (GPT-4o mini) Our platform fee Your total
500 ~$524 ~$1 $0 (free tier) ~$1
1,000 ~$1,019 ~$3 $29 ~$32
5,000 ~$5,029 ~$15 $29 ~$44
10,000 ~$9,929 ~$30 $29 ~$59

API cost estimates based on GPT-4o mini at ~600 tokens per conversation ($0.15/1M input, $0.60/1M output). Actual costs vary by conversation length and model.

See exactly what you'd pay.

SupportKit Intercom Fin Zendesk AI Chatbase
Monthly platform fee $0–$29 $29–$132/seat $55–$115/seat $19–$99
AI cost model Your API key (wholesale) $0.99 per resolution ~$1.50 per resolution Your API key
1,000 conversations ~$32 ~$1,019 ~$1,605 ~$22 + plan
10,000 conversations ~$59 ~$9,929 ~$15,150 ~$119 + plan
Full CSS customization No iframe iframe-locked iframe-locked Basic only
White-label (no branding) $29/mo $132/mo Enterprise only $199/mo
Native tool-calling Webhooks + MCP Proprietary ($89+)
Multi-provider BYOK OpenAI, Anthropic, Gemini OpenAI only
Open-source widget MIT license

Intercom costs: $29/seat (Starter) + $0.99/resolved conversation. Zendesk costs: $55/seat (Suite Team) + $50/seat Advanced AI + $1.50/automated resolution. Chatbase BYOK available on Hobby ($19/mo) with 2,000 message limit; unlimited on Standard ($99/mo). Our costs: API costs paid directly to your provider.

Built for developers who want control.

CSS

Widget that actually fits your brand

No iframe. No z-index: 9999. No "Powered by [Someone Else]" badge unless you want it. The widget mounts directly into your DOM — it inherits your fonts, respects your spacing, and switches between your light and dark modes. Style every element with CSS custom properties.

TOOLS

AI that takes action, not just answers

Define tools with a JSON schema and a webhook URL. When a customer asks to cancel their subscription, check an order status, or update their email, the AI calls your API directly — no Zapier middleware, no copy-pasting ticket IDs. Works with OpenAI function-calling and Anthropic tool_use out of the box.

BYOK

Your keys, your data, your cost

Connect your own OpenAI, Anthropic, or Google Gemini API key. We never see your conversations at the LLM layer. Your data agreement is directly with your AI provider — useful if you need zero-data-retention or are working toward HIPAA/SOC2 compliance.

EDGE

Runs at the edge

Built entirely on Cloudflare Workers, D1, and R2. The widget loads from a global CDN. Conversations are processed milliseconds from your users. Infrastructure costs us under $20/month — which is why we can charge you $29.

Simple pricing. No per-resolution fees. Ever.

Free

$0/mo

Get started with AI support and see the savings for yourself.

  • 1 widget
  • 500 conversations/month
  • BYOK (bring your own AI key)
  • "Powered by SupportKit" badge
  • Basic CSS customization
Join waitlist

Agency

$149/mo

Multi-brand management for agencies and white-label deployments.

  • Everything in Pro
  • Unlimited sub-accounts (brands)
  • Custom domains per brand
  • Isolated configs, knowledge bases, and tools per brand
  • Priority support
Join waitlist

All plans: BYOK required on Free. Optional on Pro/Agency — use your own key to cut costs, or upgrade to a managed option coming soon.

Common questions.

Why do I need to bring my own API key?
So you pay OpenAI or Anthropic directly — at their published rates — with no markup from us. On GPT-4o mini, a typical support conversation costs about half a cent. We make money from the platform subscription, not from reselling inference.
What if I don't want to manage an API key?
A managed option (we handle the key, you pay a small per-conversation fee) is on the roadmap. For now, getting an OpenAI key takes about 3 minutes at platform.openai.com.
How is this different from Chatbase?
Chatbase supports BYOK but offers limited CSS control, no native tool-calling (your AI can answer questions but can't take actions), and charges $199/month to remove their branding. We support full CSS customization, webhook and MCP tool-calling, and branding removal at $29/month.
What does "no iframe" mean?
Intercom and Zendesk embed their widgets inside iframes — a separate browser context that prevents your CSS from touching the widget. Our widget mounts directly into your page's DOM, so it inherits your fonts, can be styled like any other element, and doesn't fight your layout.
Is it really open-source?
The widget UI component is MIT-licensed and published on GitHub. The backend platform (the part that handles conversations, knowledge bases, and billing) is hosted — not open-source. You can self-host the widget component and point it at your own backend if you want full control.
What AI providers are supported?
OpenAI (GPT-4o, GPT-4o mini) and Anthropic (Claude Sonnet, Claude Haiku) at launch. Google Gemini and Mistral on the short-term roadmap.