BYOK Explained: Why Bring Your Own Key Changes Everything

The Message Cap Problem

Every AI chatbot platform charges based on messages. Chatbase: 20-40,000 credits per month. DocsBot: limited by source pages. SiteGPT: capped by plan tier. The more your users ask, the more you pay.

This creates a perverse incentive: you want your chatbot to be useful (more conversations = more deflected tickets), but every conversation costs you money. Success is punished with higher bills.

For indie makers and small teams, this is especially painful. You can't predict support volume — it spikes around launches, feature releases, and viral moments. A message cap means your chatbot might go silent exactly when you need it most.

What Is BYOK?

BYOK stands for Bring Your Own Key. Instead of paying LaunchChat for AI usage, you bring your own API key from an AI provider (OpenAI, Anthropic, or OpenRouter) and pay them directly.

LaunchChat charges a flat $19/month platform fee for BYOK. That covers:

• The RAG pipeline (chunking, embedding, retrieval)
• Widget hosting and delivery
• Analytics dashboard
• Knowledge gap tracking
• Auto-escalation
• All platform features

The AI inference cost — the actual LLM call that generates each answer — goes directly to your AI provider at their published rates. No markup from LaunchChat.

The Economics

Let's do the math for a typical support chatbot:

Cost Per Conversation

A typical support conversation involves:

• User question: ~50 tokens
• Retrieved context (3-5 chunks): ~1,500 tokens
• System prompt: ~200 tokens
• AI response: ~300 tokens
• Total: ~2,000 tokens per turn

Most conversations are 1-3 turns. Let's say 2 turns average = ~4,000 tokens.

With Claude 3 Haiku (the default model):

• Input: $0.25 per 1M tokens
• Output: $1.25 per 1M tokens
• Cost per conversation: ~$0.02-0.05

With GPT-4o-mini:

• Input: $0.15 per 1M tokens
• Output: $0.60 per 1M tokens
• Cost per conversation: ~$0.01-0.03

Monthly Cost at Scale

Add the $19/month platform fee, and you're looking at $21-69/month for 1,000 conversations. Compare that to Chatbase's $99/month Standard plan (2,000 credits) or SiteGPT's $49/month (limited chatbots).

The key difference: with BYOK, there's no cap. If you get 10,000 conversations in a month, you pay for 10,000 conversations. No throttling, no "upgrade to continue" messages, no silent failures.

How to Set Up BYOK

Step 1: Get an API Key

Choose your provider:

OpenRouter (recommended): Gives you access to all major models (Claude, GPT-4, Llama, Mixtral) through a single API key. Sign up at openrouter.ai and create an API key.

OpenAI: Sign up at platform.openai.com, go to API Keys, and create a new key.

Anthropic: Sign up at console.anthropic.com and create an API key.

Step 2: Add Your Key to LaunchChat

In your LaunchChat dashboard:

1. Go to Widget Settings → AI Model
2. Select "Bring Your Own Key"
3. Paste your API key
4. Choose your preferred model

Your key is encrypted and stored securely. LaunchChat uses it only to make inference calls on your behalf.

Step 3: Choose Your Model

With OpenRouter, you can use any available model:

• Claude 3 Haiku: Best balance of speed, quality, and cost for support
• Claude 3.5 Sonnet: Higher quality for complex technical docs
• GPT-4o-mini: Cheapest option with good quality
• GPT-4o: Premium quality, higher cost
• Llama 3 70B: Good quality, very low cost
• Mixtral 8x7B: Fast and affordable

You can change models anytime. Different widgets can use different models.

When BYOK Makes Sense

BYOK is ideal when:

• Unpredictable volume: You can't predict how many conversations you'll have
• High volume: More than 2,000 conversations per month
• Cost sensitivity: You want to pay exactly what you use, no more
• Model flexibility: You want to choose or switch AI models
• Launch spikes: You expect traffic surges around launches or marketing campaigns

BYOK is less ideal when:

• Low volume: Under 500 conversations per month (the included messages on Pro are sufficient)
• Simplicity: You don't want to manage an API key
• Budget certainty: You prefer a fixed monthly cost with no variables

BYOK + Lifetime Deal

Here's the power combo: buy the lifetime deal ($229 one-time) for the platform, then use BYOK for unlimited messages.

Your total ongoing cost: just the AI API fees (~$0.02-0.05 per conversation). No monthly platform fee. No message caps. No recurring charges.

For an indie maker handling 500 conversations per month, that's roughly $10-25/month in API costs — with zero platform fees after the initial $229.

Security

Your API key is:

• Encrypted at rest using AES-256
• Never exposed in client-side code
• Used only for server-side inference calls
• Deletable at any time from your dashboard

LaunchChat never stores your API key in logs, analytics, or any other system. It's used exclusively for making LLM calls on your behalf.

Getting Started

1. Sign up for LaunchChat (free tier to test)
2. Connect your documentation
3. Go to Widget Settings → AI Model
4. Select BYOK and paste your API key
5. Choose your model
6. Done — unlimited messages at API cost

The free tier lets you test the full pipeline before committing to BYOK. Once you're satisfied with answer quality, switch to BYOK and remove the message cap entirely.

Start free or see pricing for all plan options.