The stack under momor.ai, opening to developers.
We run a judgment-aware orchestration engine and a multi-provider model gateway in production every day. They power Momor Search and our enterprise deployments. We are packaging both for developers. This page represents the honest state of that work.
We are a small company, and this is what we're building.
No vaporware, no empty docs portal. The engine and the gateway are real — they have run our public product since launch. The developer packaging (APIs, SDK surface, documentation) is being built right now by a small team that ships. If you want in early, talk to us and help shape it.
Momor Model Gateway
One request, every model.
A single request shape that runs identically on every configured provider, with mid-conversation failover, two-layer caching, and per-call cost accounting — the layer that keeps momor.ai answering when providers stumble.
Momor Workflows
Compose the process. Momor runs it.
A workflow engine for AI-driven processes: compose models, tools, and human checkpoints into a running process — executed, recorded, replayable, and never losing its place, whether the work takes seconds or months.
The orchestration stack for .NET teams.
The agent-framework wave is Python-first. Momor's stack is .NET-native — built in .NET, run in production on .NET, and packaged for the teams the ecosystem keeps skipping.
It runs in production, in public, every day.
Momor Search at momor.ai is powered by this exact stack — the same engine, the same gateway, under real traffic. So is ArtsyTells, a drawing-prompt site for kids, at the other end of the workflow spectrum. Try either before you talk to us.
Want in early?
Early access means shaping the SDK and Gateway with us while they're built. Tell us what you're building.