01
Create the agent
Start with the job: receptionist, support, lead qualifier, or a focused sub-agent.
How it works
Match the model to the work
- 02
Choose a model
Browse the catalogue with each model's price shown in AI credits per 1,000 tokens, or switch to your own OpenRouter or Vercel AI Gateway key.
- 03
Tune as you learn
If answers are thin, move to a stronger model. If simple FAQs cost too much, move to a lighter one. No rebuild.
- 04
Mix across sub-agents
The main agent and each of its sub-agents can run on whichever model suits the job.
Capabilities
Why model choice belongs in a no-code builder
You shouldn't need an engineering team to change models.
A large catalogue
200+ models from OpenAI, Anthropic, Google, Mistral, Meta, Qwen, DeepSeek and more. Every one supports the tool calls and structured replies an agent needs.
Bring your own key
Add an OpenRouter or Vercel AI Gateway key to an agent. Its chat and voice replies are then billed to your account, not your Embage AI credits.
Keys stay encrypted
Your key is encrypted before it's stored and is never sent back to the dashboard after you save it.
Per agent
A support agent can run a stronger model while a simple receptionist stays light and fast.
Per sub-agent
A knowledge search sub-agent and a post-conversation reviewer don't have to share a model.
Swap without rebuilds
Change the model in settings. Knowledge, forms and tools stay as they are.
In practice
Simple rules of thumb
Start practical, then adjust based on real conversations.
Receptionist and routing
Lighter, faster models suit jobs like greeting, sorting and handing off.
Deep support
Stronger models suit questions that need policy detail or multi-step troubleshooting.
Structured extraction
Post-conversation reviewers work best on a model that fills in form fields carefully.
High-volume FAQs
If most chats are short and answered from your knowledge base, keep the main model lean.
FAQ
Questions
Which providers can I bring?
OpenRouter and Vercel AI Gateway. You still pick the model from the Embage catalogue, so every option supports what an agent needs.
What does my own key cover?
The agent's own replies in website chat, Telegram and voice, plus the default summary that runs after each conversation. Sub-agents pick their own models and run on Embage AI credits.
What still counts toward my plan?
Voice sessions, knowledge searches, datastore writes and integration calls have monthly allowances on every plan, whichever key you use. Speech recognition and text-to-speech for voice are provided by Embage.
Is there a free plan if I bring my own key?
Yes. The Free plan is enough for a website agent when you add your own OpenRouter or Vercel AI Gateway key: chat replies use 0 Embage AI credits, with the free monthly allowances for search, datastore, and integrations. Paid plans include Embage credits when you want Embage to cover model spend. Voice sessions start on Starter.
Do I lock into one model forever?
No. You can change models as traffic, quality or budget changes, without rebuilding the agent.
How does this affect pricing?
Paid plans include 1,000 (Starter) or 3,800 (Pro) AI credits a month, and model prices are shown in credits. A lighter model uses fewer credits per message, and your own key uses none for the agent's replies.
Free plan · paid plans free for 7 days
Build your first customer agent.
No credit card. Questions? Write to support@embage.com.