A customer platform,
built like infrastructure
A streaming pipeline instead of nightly batch syncs, at-most-once sends, one SDK with a clean event model, and a data plane you can run in your own VPC.
A streaming spine,
not a nightly batch
Events move through the pipeline over NATS JetStream, land in ClickHouse for analytics and never wait for a sync window. Every send is claimed before the provider call, so a crash or a deploy can’t send it twice. If the outcome is unknown, it’s recorded and not re-sent.
Fix the cause,
then replay what failed
Events that can’t be processed park in a dead-letter queue with their error. See what’s parked, fix the cause, and send them through again. Replay always starts with a dry run that changes nothing.
# what is parked, by error code $ origamy dlq # fix the cause first $ origamy doctor # a dry run, then it asks before sending $ origamy dlq replay
Your VPC, your data,
one chart
The data plane installs into your EKS, GKE or AKS cluster from a single Helm chart: gateway, streams, ClickHouse, identity and the rest. It reaches the control plane over an outbound-only tunnel, and your customer data never leaves your network.
Deployment guide in the docs# the whole data plane, one chart $ helm install odp \ oci://ghcr.io/qubelylabs/charts/origamy-data-plane # NATS, Redis, ClickHouse, Postgres run inside it # or let the CLI run the install for you $ origamy deploy $ origamy status
Your AI tools
can build here too
A remote MCP server with OAuth sign-in. Connect Claude Code, Cursor or another MCP app, then plan segments, templates, forms and journeys from where you already work. Drafts land off, and people’s data never comes back.
How to connect$ claude mcp add --transport http origamy \
https://mcp.origamy.io/mcp
# then /mcp to sign in and pick a workspace
origamy