Never billed per token
Bring your own LLM key and messages are unlimited on every plan — inference runs at provider cost with zero token markup. Cloud plans are priced by monthly active users (MAU), not tokens — and self-hosting stays free.
Put a real agent in front of real users on managed Cloud.
No credit card required.
For products with a growing user base.
Higher limits for teams shipping at volume.
Security requirements, custom capacity, or a delivered production agent.
The complete self-hosted agent runtime. No limits from us.
MAU, explained
A monthly active user is a unique person who chats with your agent at least once in a billing period. In practice, MAU roughly tracks conversations — not messages.
For self-hosted teams at scale
Community Edition already includes the audit log. Enterprise Edition layers paid operations features on top of a self-hosted install — activated with a license.jwt, no reinstall. Two features are available with an EE license today; the rest are in development; release timing will be announced.
Want to know when the full Enterprise Edition ships? Join the waitlist.
Pricing questions
Cloud plans are priced by monthly active users (MAU), plus schema and knowledge-base document limits per tier. You are never billed per message or per token: with your own LLM key, messages are unlimited on every plan; the free demo key includes 50 messages per month.
A monthly active user is a unique person who chats with your agent at least once in a billing period — roughly, it counts conversations, not messages. One user can send hundreds of messages and still counts once. Anonymous widget visitors are counted once using a persistent visitor id.
Yes. BYOK — bring your own key — is included everywhere, in both Community Edition and Cloud, with zero token markup: inference runs at your provider’s list price. Configure any provider key in the Admin Dashboard or pass it per request.
Limits, attribution, and support. Free covers up to 100 monthly active users, 2 schemas, and 25 knowledge-base documents, with a “Powered by SyntheticBrew” badge on the widget. Paid plans raise MAU, schema, and document limits, remove the attribution badge, and add email support on Pro or priority support on Business.
On the Free plan, new users are blocked once you hit the MAU limit while existing sessions continue working. On paid plans, limits are soft: you see an upgrade prompt and nothing shuts off mid-conversation. With your own LLM key, messages are never capped on any plan.
Your account downgrades to the Free plan. Your agents, schemas, and data remain — you only lose capacity above the Free plan limits.
Yes. The self-hosted Community Edition contains the full agent runtime and is free with no time limit under BSL 1.1, automatically converting to Apache 2.0 after four years. Community Edition features will not move behind a paywall.
Yes. Community Edition self-hosts with Docker Compose, Kubernetes using the Helm chart, or a standalone binary alongside PostgreSQL. There is no cloud dependency and no phone-home requirement.
OpenAI, Anthropic, Google Gemini, Azure OpenAI, Ollama, OpenRouter, and any OpenAI-compatible provider such as DeepSeek or Mistral. The Model Registry ships with 14+ models and tier recommendations, and you can add custom model configurations.
Enterprise Edition activates on top of a self-hosted install: you download a license.jwt from your dashboard, place it at ~/.syntheticbrew/license.jwt, and restart the Engine. Without a license the Engine keeps running as Community Edition — nothing breaks.
The engine is open source — inspect it, self-host it free, and pay only when Cloud capacity or enterprise requirements actually matter to you.