Use agents your way.
Free to try, bring your own key to scale it up, or run the open-source engine yourself — no markup on model calls, ever.
Frequently asked questions
Why is the visual builder only on Enterprise or self-hosted?
Chat is cheap for us to run — it's a pass-through to a model. The builder deploys real, running workflow instances, which is genuine infrastructure to operate at scale, and we're not yet set up to run that for free or Explorer accounts. Enterprise gets it hosted as part of a direct conversation about what you need; self-hosting gets it because you're running (and paying for) your own infrastructure.
Why does Explorer cost anything if I bring my own key?
The $5/mo covers our own infrastructure — file parsing, embeddings, storage, and request processing. Your model calls are billed directly to your own OpenAI/Anthropic/Gemini/OpenRouter account, never marked up by us.
What happens when I hit the daily message limit?
Requests are rate-limited, not silently dropped — you get a clear response telling you when the limit resets. Connect your own key (Explorer) to raise it from 5/day to 100/day, or self-host for no cap at all.
What's the actual difference between Explorer and self-hosting?
Explorer is us running the chat platform for you — you just supply the model key. Self-hosting means you run Agent Substrate and the chat UI on your own infrastructure, with no relationship to this platform at all, and you get the builder too. Same engine, either way.
How do I get Enterprise limits?
There's no self-serve signup for it yet — reach out and we'll set up a plan that matches what you actually need.