Connect AI (Bring Your Own Key)
Bring your own API key and Serenities calls that AI directly — in the builder chat, your sites, and flows. You pay your provider at their rates; we add no AI markup.
This page is about bringing AI *into* Serenities. You add a provider key and Serenities uses that model.
Want the opposite — to drive Serenities from Claude or another AI tool? That’s MCP: it connects Serenities to an outside tool so the tool can use your databases, sites and flows. See Connect Claude via MCP.
No AI markup
You pay your AI provider directly at their rates. We don’t charge extra for AI usage.
Your keys, your control
Keys are encrypted at rest and shown masked. Test, disable, or delete any key anytime.
Use what you have
Already pay for OpenAI or Claude? Use that same account here — no double paying.
Providers you can connect
Pay-per-token API keys — billed by the provider as you use them.
OpenAI
GPT models in the builder + voice transcription (pay-per-token on your OpenAI account)
Key looks like sk-proj-...
Anthropic (Claude)
Claude Opus/Sonnet in the builder (pay-per-token on your Anthropic account)
Key looks like sk-ant-api03-...
Google (Gemini)
Gemini in the builder (pay-per-token on your Google AI Studio account)
Key looks like AIza...
Azure OpenAI
GPT models running under your own Azure subscription and data rules. Add each deployment name as a model.
Key looks like Azure OpenAI API key
Amazon Bedrock
Claude, Nova, Llama and DeepSeek through your AWS account (Bedrock model access must be enabled in that region).
Key looks like AWS secret access key
xAI (Grok)
Grok in the builder (pay-per-token on your xAI account)
Key looks like xai-...
OpenRouter
One key for 300+ models (Claude, GPT, Gemini, Llama, Qwen…) — browse the live list below
Key looks like sk-or-v1-...
DeepSeek
DeepSeek chat and reasoner models (pay-per-token on your DeepSeek account)
Key looks like sk-...
Mistral AI
Mistral Large, Codestral, Devstral (pay-per-token on your Mistral account)
Key looks like ...
Groq
Very fast open-weight models (Llama, Qwen, GPT-OSS) on Groq hardware
Key looks like gsk_...
Together AI
Open-weight models (Llama, Qwen, DeepSeek, Kimi) hosted by Together
Key looks like ...
Fireworks AI
Open-weight models hosted by Fireworks
Key looks like fw_...
Cerebras
Ultra-fast inference for open-weight models
Key looks like csk-...
Perplexity
Sonar models with built-in web search (best for research chats, limited tool use)
Key looks like pplx-...
Cohere
Command models via Cohere’s OpenAI-compatible endpoint
Key looks like ...
Moonshot (Kimi API)
Kimi models pay-per-token (not the coding plan)
Key looks like sk-...
Z.ai (GLM API)
GLM models pay-per-token (not the coding plan)
Key looks like ...
MiniMax (API)
MiniMax models pay-per-token (not the coding plan)
Key looks like eyJ...
Alibaba (Qwen)
Qwen models via Alibaba Model Studio (international endpoint)
Key looks like sk-...
DeepInfra
Open models (Llama, DeepSeek, Qwen) served cheaply, pay-per-token
Key looks like DeepInfra API key
Hugging Face
Hugging Face Inference Providers — thousands of open models behind one token
Key looks like hf_...
Baseten
Baseten Model APIs — fast hosted open models (DeepSeek, Kimi, GPT-OSS)
Key looks like Baseten API key
SambaNova
Very fast inference for open models (Llama, DeepSeek, Qwen)
Key looks like SambaNova API key
Cloudflare Workers AI
Open models on Cloudflare's network, billed to your Cloudflare account
Key looks like Cloudflare API token (Workers AI: Read)
Vercel AI Gateway
One key for every model on Vercel’s AI Gateway
Key looks like vck_...
Coding subscription plans
Flat-rate plans — connect the plan token instead of paying per token.
How to connect a key
Get an API key from the provider
Use the “Get key” link on any provider above, sign in, and create a key.
Add it under Your AI (left rail) → pick the provider → Add key
Give the key a name — you can keep several per provider (“Work”, “Personal”) and edit one later to rename it or paste a new secret without touching anything that uses it. Serenities tests the key before saving.
Pick that model in chat, in an AI agent, or from a function
Every chat’s model menu lists your keys with their models; agents offer the same list; functions call ctx.ai.generate with the model id. Switch a key off or delete it anytime.