Service Keys — LLM, TTS, STT Inference Access

Service Keys (starting with <code>oss_sk_</code>) authenticate your workflows to Shopbell's managed Model Proxy Service. The Model Proxy hosts LLM inference (GPT-4o, Claude, OSS Llama), text-to-speech (Kokoro), and speech-to-text (Deepgram). When you select 'Shopbell-managed' for any model in a workflow, that workflow uses your service key to bill inference costs to your plan. This guide covers how to get service keys and what they cost.

Your default service key

When you sign up, Shopbell creates a Default Shopbell Model Service Key for you. It's already attached to your first workflow's model configuration. Find it in Developer Portal under 'Shopbell Service Keys'. It's labeled 'Active' and has a 'Created' date from your signup day.

Generating additional service keys

Click 'New Service Key'. Name it ('Production', 'Testing', 'Compliance-Reviewed', etc.). A new oss_sk_ key is generated. Attach the new key to specific model configurations. Useful when you want to track usage per agent or per environment — each service key's usage appears separately in the Usage page.

Note: in some self-hosted deployments, the link says 'To generate additional service keys, Sign up on app.shopbell.ai' with a clickable link — that link now correctly points to our app (post Phase-1 rebrand). On the Shopbell hosted plan, you can generate service keys directly without an external signup.

Billing model — flat $99/mo, no per-minute charges

Shopbell operates on a flat $99/month subscription. Service key usage (LLM tokens, TTS characters, STT minutes) is included in your subscription. There are no per-minute charges, no per-call charges, no credit purchases needed. If your usage exceeds typical small-business volume (typically 5000+ minutes/month), our team will reach out to discuss a custom plan — but you will never be silently billed for overages or have service interrupted.

Frequently Asked Questions