A Cloudflare-native platform, isolated per tenant by design.
For CTOs, architects, and technically sophisticated merchants: here is how Compass is actually built — the compute model, the per-tenant data isolation, the event spine, and the abstractions that let it evolve.
Edge compute with per-tenant databases
Compass runs on Cloudflare Workers with a database per tenant, so isolation and scale are structural rather than configured after the fact.
Cloudflare Workers
Application logic runs on Cloudflare Workers at the edge — no servers to provision, low latency close to your customers, and horizontal scale by default.
D1 tenant databases
Each tenant gets its own D1 SQL database. Isolation is a property of the storage layer, and per-tenant databases keep blast radius small.
Global control plane
A control plane coordinates tenants, routing, and provisioning across the network, while tenant data stays in its own database.
Durable Objects
Durable Objects provide single-threaded, consistent coordination points for stateful workflows that need strong ordering.
A durable, validated event spine
Events are the primary abstraction. Queues, Workflows, and provider adapters move them from source to profile reliably.
Queues
Cloudflare Queues decouple ingestion from processing, so bursts of events are absorbed and handled reliably rather than dropped.
Workflows
Durable Workflows orchestrate multi-step processes — like ingestion, resolution, and fan-out — with retries and resumability.
Event spine
A validated event spine is the backbone of the platform: events flow through a common envelope onto the customer timeline.
Provider adapters
Adapters normalize third-party providers — commerce, email, surveys — into the same event model, so new sources plug in cleanly.
The right store for each job
R2, KV, and D1 each handle what they are best at, behind a storage-driver abstraction that keeps the platform portable.
R2 object storage
R2 holds larger objects and assets without egress penalties, keeping media and exports close to the compute that uses them.
KV
Workers KV provides fast, globally distributed reads for configuration and lookups that need to be everywhere at once.
AI Gateway
AI calls are routed through an AI Gateway, giving a controlled, observable path for model usage across the platform.
Storage-driver abstraction
A storage-driver abstraction sits between the app and its stores, so the underlying engine can evolve without rewriting the platform.
Isolation today, a clear path to scale tomorrow.
Because each tenant is its own database behind a storage-driver abstraction, Compass has a defined tenant migration path. Tenants can be moved or upgraded as they grow, without redesigning the application — the abstraction absorbs the change.
How tenant isolation works- One database per tenant, isolated by design
- Storage-driver abstraction decouples engine from app
- A defined path to migrate and scale tenants
- Global control plane coordinates provisioning
Architecture that makes the product’s promises true.
Provable audiences, enforced consent, and trustworthy AI are only as good as the foundation under them. Per-tenant isolation, a validated event spine, and a controlled AI Gateway are what let Compass make those claims with a straight face.
- Isolation backs the security and privacy story
- The event spine backs profile auditability
- The AI Gateway backs grounded, observable AI
- Adapters make new integrations low-risk
Building on Cloudflare gives Compass a globally distributed, horizontally scalable foundation without operating fleets of servers. If your team wants to go deeper on any layer, we are happy to walk through it.
Bring your architecture questions.
We will walk your team through the compute model, isolation guarantees, and the event pipeline in as much detail as you need.