Pricing & limits
SyntheticBrew Cloud is priced by monthly active users (MAU) — the number of unique people who chatted with your agent in the last 30 days. Messages are never metered: you can send as many as you like on every plan. Heavy inference is covered by your own LLM key at your provider’s cost, with no token markup from us.
The self-hosted Community Edition is free forever and has no MAU, agent, or document limits — you bring your own hardware and LLM keys.
| Plan | Monthly active users | Deployed agents | Knowledge-base documents | Support |
|---|---|---|---|---|
| Free | 100 | 2 | 25 | Community |
| Pro | 2,000 | 3 | 100 | |
| Business | 20,000 | 25 | 500 | Priority |
| Enterprise | Custom | Custom | Custom | Dedicated onboarding |
Pro is $49/mo ($490/yr — two months free). Business is $199/mo ($1,990/yr). Enterprise is quoted per deployment — contact us.
What counts as an active user
Section titled “What counts as an active user”A monthly active user (MAU) is a unique end-user who sent at least one message to your agent within the last 30 days. This is a rolling window, not a calendar month or billing period — there is no reset day: a user counts as active for exactly 30 days after their last message and then ages out. A single user is counted once no matter how many messages they send, so a chatty user costs the same as a one-message user.
- Authenticated end-users are identified by their user id.
- Anonymous widget visitors are counted once each via a persistent visitor id stored in the browser, so a returning visitor within the same window is not double-counted.
MAU is best read as “roughly the number of conversations, not the number of messages.”
How limits behave
Section titled “How limits behave”Limit enforcement differs by plan so that a free deployment can’t accidentally exceed its allowance while a paying customer is never cut off mid-service:
- Free plan — hard cap. Once your rolling 30-day count reaches 100 active users, new users are blocked from starting a chat until the count drops back below the limit (users age out of the window 30 days after their last message). Users who are already chatting keep their sessions; existing conversations are never interrupted.
- Paid plans — soft cap. Service continues past the included MAU allowance — your users are never blocked mid-period, and overage is never billed automatically. Exceeding the allowance only triggers an upgrade prompt in the dashboard; if your traffic has outgrown the plan, move up a tier or contact us for a custom allowance.
The agent and knowledge-base document limits are provisioning limits: you can create up to the plan’s number of deployed agents and stored documents. Messages and reasoning steps are not capped on any plan.
Bring your own key (BYOK) — everywhere
Section titled “Bring your own key (BYOK) — everywhere”Every plan, including Free and self-hosted CE, supports BYOK. Configure any provider’s API key in the Admin Dashboard, or pass it per request with the BYOK headers. Inference then runs entirely on your key at your provider’s list price — SyntheticBrew never adds a token markup.
Every Cloud plan also ships a shared demo LLM key so you can onboard and try an agent before wiring up your own provider. The demo key is limited to 50 turns per 30-day window and is intended for evaluation, not production traffic — switch to BYOK for anything real.
Attribution on free plans
Section titled “Attribution on free plans”Free-plan agents include a “Powered by SyntheticBrew” attribution:
- the agent is instructed via its system prompt to identify that it’s powered by SyntheticBrew — best-effort, since how consistently the agent follows the instruction depends on the model, and
- a deterministic “Powered by SyntheticBrew” badge is shown on the embeddable chat widget.
Upgrading to Pro or Business removes both the system-prompt instruction and the widget badge. Enterprise plans are unbranded by default.
What’s next
Section titled “What’s next”- BYOK — bring your own model key on any plan
- Usage limits — self-hosted, operator-configurable turn/step quotas
- Docker deployment — run the free self-hosted Community Edition