Head-to-head

SyntheticBrew vs Chatbase: rent a support bot, or own one?

Both put a grounded AI support agent on your website. Chatbase is a polished closed SaaS billed in message credits. SyntheticBrew is a source-available engine you can self-host or run in our cloud — pay your LLM provider directly, keep your data, and set the whole thing up from Cursor or Claude Code via native MCP. All Chatbase figures below are from their public pricing and docs, July 2026.

Side by side

SyntheticBrew vs Chatbase: side-by-side comparison

Dimension	SyntheticBrew	Chatbase
What it is	Source-available agent platform — self-host it or use our cloud; agents ship as an embeddable website widget	Closed SaaS for AI customer-support agents, managed entirely on Chatbase infrastructure
Pricing model	Self-hosted is free — you pay only your LLM provider. Cloud has a free tier and flat plans	Message credits: one reply costs 1–5 credits depending on the model, and enabled AI Actions multiply requests per reply
Source available / self-hosting	BSL 1.1 engine on GitHub; Docker and Helm deployment	Closed source, cloud only
Bring your own LLM key	Yes — add models on your own provider keys (any OpenAI-compatible endpoint or OpenRouter); plus an optional per-request end-user BYOK mode admins can enable	No documented bring-your-own-key option; usage runs on pooled message credits
Manage from a coding agent (MCP)	Native MCP server in the official registry: connect Cursor or Claude Code over OAuth and provision agents, knowledge, and the embed snippet from your editor	No MCP support documented; third-party wrappers over the REST API only
API access	Full REST + SSE API on every tier, including free and self-hosted	API access starts on the Standard plan ($120/mo billed annually)
Agents included	Multiple agents with runtime delegation between them	1 agent on every self-serve plan; each extra agent is $300/year
Remove vendor branding	Self-hosted: it is your deployment — nothing to remove	"Powered by Chatbase" removal is a $1,188/year add-on (free on Enterprise)
Built-in channels	Website widget + REST/SSE API today; other channels bridge through your backend	Broad built-in catalog: WhatsApp, Messenger, Instagram, Slack, Zendesk, Salesforce, Intercom, and more
Voice & telephony	Not offered today	Included from the Standard plan
Knowledge grounding	Knowledge bases with RAG plus typed knowledge graphs; grounded, cited answers	Training on sites, files, and text (1–40 MB depending on plan)
Data ownership	Self-hosted: agents, knowledge, and history live in your PostgreSQL	Vendor-managed; free-plan agents are deleted after 14 days of inactivity

Where SyntheticBrew differs

Where SyntheticBrew differs from Chatbase — and why it matters

Token prices, not credit arithmetic

Chatbase meters in message credits: a single reply costs 1 to 5 credits depending on the model, AI Actions multiply the requests behind one answer, and refills are $40 per 1,000 credits. SyntheticBrew has no credit layer — bring your own LLM key and pay your provider’s raw token price, or self-host for free.

Why it matters: Credit systems make support costs hard to predict: switching to a better model or enabling an action can double or triple burn overnight. Paying token prices directly keeps the bill proportional to actual usage — and lets you choose models on merit, not on credit multipliers.

Set up from Cursor, not from a dashboard

SyntheticBrew ships a native MCP server, listed in the official MCP registry. Connect it to Cursor, Claude Code, or any MCP client over OAuth, and your coding agent provisions the agent, loads knowledge, and hands you the embed snippet — one prompt, end to end. Chatbase documents no MCP support.

Why it matters: If your team already lives in a coding agent, the fastest setup is the one you never leave your editor for. It also means your agent config is reproducible — re-run the prompt, get the same deployment.

An exit that costs nothing

The engine is source-available (BSL 1.1) and self-hostable: your agents, knowledge bases, and conversation history live in your own PostgreSQL. On Chatbase, everything stays on their infrastructure, and free-plan agents are deleted after 14 days of inactivity.

Why it matters: A support agent accumulates real operational knowledge. Owning the deployment means a pricing change, an acquisition, or a product pivot on the vendor’s side never holds your support channel hostage.

For enterprise teams

An enterprise-ready Chatbase alternative

Runs entirely inside your perimeter: engine + PostgreSQL, air-gap friendly, bring your own model keys.
Ed25519-signed JWTs, scoped API tokens, and an audit trail of agent actions.
No per-seat or per-agent surcharges on self-hosted deployments.
Standard BSL 1.1 license with a fixed Apache 2.0 conversion date — read the code before you commit.

Enterprise AI agents

Choose for the job

Pick by what you value

Need WhatsApp, Instagram, or voice channels out of the box with zero infrastructure? Chatbase covers that today. Want predictable token-price economics, a deployment you own, an API from the free tier, and setup driven by your coding agent? That is exactly what SyntheticBrew is built for — and you can read the engine's source before deciding.

Try Cloud Self-host Contact sales

Questions

SyntheticBrew vs Chatbase: FAQ

Is SyntheticBrew a good Chatbase alternative?

If you want to own the deployment, pay LLM providers directly instead of buying message credits, or drive setup from Cursor / Claude Code via MCP — yes. If you need built-in WhatsApp, Messenger, or voice channels with zero infrastructure work, Chatbase currently covers more channels out of the box.

When is Chatbase the better choice?

When nobody on the team will ever touch a terminal and you need built-in messaging channels (WhatsApp, Instagram, Messenger, Slack) or telephony from day one. Chatbase’s fully managed flow has less friction for that profile — at the cost of credits-based pricing and no self-hosting.

What does each actually cost to run?

SyntheticBrew self-hosted is free under BSL 1.1 — you pay only your LLM provider’s token prices; the cloud has a free tier. Chatbase plans run $32–$400/month (billed annually) with 500–15,000 message credits, where one reply burns 1–5 credits by model; extras include $40 per 1,000 credits, $300/year per additional agent, and $1,188/year to remove their branding.

Can I migrate from Chatbase to SyntheticBrew?

Your knowledge sources are your own docs and site content — re-ingesting them into a SyntheticBrew knowledge base takes minutes, from the dashboard or via MCP from your editor. There is no automated importer, but there is also nothing proprietary to convert.

Which LLMs can I use with each?

SyntheticBrew works with any OpenAI-compatible endpoint or OpenRouter — hundreds of models, on your key, switchable per agent, with an optional per-request end-user BYOK mode. Chatbase offers a curated model list (OpenAI, Anthropic, Gemini, and others) running on their credits, with higher-tier models consuming more credits per message.

Is SyntheticBrew really open source?

The engine is source-available under BSL 1.1 — free to self-host, including multi-tenant use inside your product; the only restriction is offering the engine itself as a competing managed service. It converts to Apache 2.0 on a fixed date. We say “source-available” rather than “open source” because BSL is not an OSI-approved license.

More comparisons: SyntheticBrew vs Botpress · SyntheticBrew vs Dify · SyntheticBrew vs Flowise · SyntheticBrew vs n8n · All comparisons