Skip to content

The router takes requests, placement decides which worker owns each id, and workers run your code and keep its state. The modes differ in who runs each part, and where.

Fully managedLeast effort
Assistant CloudRouterPlacementWorkers · state
Bring your own nodesNo data at rest with us
Assistant CloudRouterPlacement
Your cloudWorkers · state
Bring your own cloudWe never see requests or data
Assistant CloudPlacement
Your cloudRouterWorkers · state
On-premiseNo production data reaches us
Your cloudRouterPlacementWorkers · state
Self-hostedFull control, no vendor
Your cloudRouterPlacementWorkers · state

Fully managed

Least effort. We run everything; your workers run on microVMs that scale from zero.

Fully managed. Your app sends requests to Pinboard in Assistant Cloud. Pinboard keeps placement in Redis and routes each request by id to your workers, which run on microVMs and save state to a managed state store.Assistant CloudYour appPinboardRedisWorkersmicroVMsState storeby id
You run
Your appWorker code
We run
PinboardRedisWorkersState store

Bring your own nodes

No data at rest with us. Your workers dial out to our router, and their state stays in your cloud.

Bring your own nodes. Pinboard and its Redis directory run in Assistant Cloud. Your workers run in your cloud and open an outbound connection to Pinboard; requests for their ids come back over that connection, so the workers need no public address or inbound port. Workers keep state in your own state store.Assistant CloudYour cloudYour appPinboardRedisWorkersState storedials out
One outbound connection per worker; no inbound ports
You run
WorkersState storeWorker code
We run
PinboardRedis

Bring your own cloud

We never see your requests or data. The data plane runs in your cloud; only placement lookups reach us.

Bring your own cloud. In your cloud, your app sends requests to a Pinboard proxy, which routes them by id to your workers; workers keep state in your state store. Assistant Cloud runs the Pinboard control plane and its Redis directory. The only traffic between the clouds is placement metadata: the proxy asks where an id lives, and workers register.Assistant CloudYour cloudPinboardcontrol planeRedisYour appPinboard proxyWorkersState storeby idwhere is id?register
The only traffic between clouds: placement metadata
You run
Your appPinboard proxyWorkersState store
We run
PinboardRedis

On-premise

No production data ever reaches us. Everything runs in your account; our operator pulls updates, and nothing connects in.

On-premise. Pinboard, Redis, workers and the state store all run in your cloud account, next to an operator we provide. The operator keeps them deployed and upgraded, and pulls its instructions from Assistant Cloud over an outbound connection; there is no inbound management connection.Your cloudAssistant CloudYour appPinboardWorkersRedisState storeOperatorUpdatesby idpulls
You run
Cloud accountWorker code
We run
PinboardRedisWorkersState store

Self-hosted

Full control, no vendor. Open source; you run all of it, from the same manifests we deploy.

Self-hosted. Everything runs in your cloud. Your app sends requests to Pinboard, which keeps placement in Redis and routes each request by id to your workers. Workers save state to your state store.Your cloudYour appPinboardRedisWorkersState storeby id
You run
PinboardRedisWorkersState store
We run
nothing

Try it on your laptop: Run a local cluster.