The open-source cloud for
AI agents and apps.
One API for 51 live models, Base backends, identity, secrets, vector search, and full-text search. Pay-as-you-go, billed per organization. Run it on Hanzo Cloud or self-host the exact same stack on your own Kubernetes.
Everything an AI product needs
Composable, open-source building blocks. Use one, use all — they work together out of the box.
51 live models, one API
OpenAI, Anthropic, and open-weight models behind a single OpenAI-compatible endpoint. Switch providers without touching code; route, fall back, and cache automatically.
Base backends
Per-organization and per-user SQLite-backed Base instances. An instant application backend with auth, records, and realtime — no database to provision.
IAM
OIDC identity for every app: SSO, OAuth2, social and Web3 login, organizations, and fine-grained access — all white-labeled to your domain.
KMS
MPC-backed key management. Store secrets once, sync them into Kubernetes as native resources. Nothing lives in plaintext, ever.
Vector search
Managed vector database for embeddings and RAG. Build semantic search and agent memory on infrastructure that scales with your index.
Full-text search
Native full-text search over your data with per-user isolation built in. No separate search cluster to run or keep in sync.
Pay only for what you use.
Every model call, every byte, every key — metered and billed per organization. No seats, no minimums, no surprise invoices. Add a balance and start shipping; share it across your whole team under one organization.
- Per-organization balances and usage records
- Transparent per-token and per-request metering
- Free tier to start — upgrade when you grow
- Self-host for $0 and bring your own provider keys
Create an organization, add a balance, and call any of 51 models. Costs accrue per use and are debited from your organization balance in real time.
Open the console Self-host it freeStart building on Hanzo Cloud
Ship your AI agent or app today. Open the console to provision models, Base, IAM, KMS, and search in minutes — or self-host the whole stack.