Query and Mutation Operations

Every unary RPC leaf and retained HTTP endpoint has query, infinite-query, and mutation builders. Your application chooses how to use each operation; neither the RPC definition nor the HTTP method determines the builder. HTTP stream builders are deferred.

Use a query when TanStack should cache a result by semantic request identity. For RPCs with a payload, the query key contains the normalized, canonical payload. TanStack can refetch or cancel the query.

Use a mutation when the call represents an action or write. Mutation variables arrive when the mutation runs, and their values do not become part of the mutation key.

Use an infinite query when TanStack should accumulate paginated results. Map each page parameter to an RPC payload or complete decoded HTTP request; the mapped initial request becomes part of the semantic key. Keep stable filters in every page request. See HTTP pagination or RPC pagination.

Use an accumulated streamed query when the application needs every emitted value in order. Use a live query when it needs only the latest emitted value. Both operations close their stream iterator when TanStack cancels the query.

const user = useQuery(rpcQuery.users.get.queryOptions({ input: { id: 1 } }))

const removeUser = useMutation(rpcQuery.users.delete.mutationOptions())
removeUser.mutate({ id: 1 })

After a successful mutation, invalidate the affected query prefix explicitly. The package does not infer relationships between API operations.

The public RPC consumer and public HTTP consumer check these builder and hook contracts.