How Ramose compares

Ramose is closest to an offline-first application database, a typed graph store, and an agent backend combined. Its local client, policy, operations, and MCP surface share one model.

The short version

If you would otherwise use	Ramose feels familiar because	The decisive difference
Convex	typed queries and server-controlled functions	complete offline replicas, your Cloudflare account, native MCP
Instant or Firebase	local-first reads and realtime synchronization	code-defined operations and policy, no hosted control plane
Supabase	the database enforces access	field-level filtered views and operation grants apply to app and agent callers
D1 + Drizzle	it runs on Cloudflare primitives	sync, history, authorization, and MCP are already assembled
Postgres + an ORM	types flow from a schema	no SQL, immutable history, offline query execution

Choose Ramose when related data belongs together

Ramose is a good fit when entities and relations should remain directly queryable under one policy and operation model.

Choose something else when SQL is the product

Ramose does not provide ad-hoc SQL, cross-database joins, a BI ecosystem, full-text search, or a managed console. If analysts need to explore the primary store directly, Postgres or a warehouse is usually the better center of gravity. Ramose data can still feed an analytical system.

The performance trade

Writes are committed in order by one authoritative writer. That gives the database simple ordering, uniqueness, receipts, and history, with a finite write ceiling.

Queries run over a complete authorized local graph in the browser and over immutable filtered values on the server. They are fast when a graph is appropriately sized and queries constrain indexed fields early. A giant graph increases initial sync, browser storage, local index, and query costs.

The maturity trade

Ramose is pre-release. APIs are intentionally changing toward the coherent model documented here. There is no hosted service, support contract, dashboard, migration framework, or broad integration catalog. In return, the whole data plane runs in your account and the application model remains ordinary TypeScript.

Next

• Mental model — see whether the architecture matches how you think
• Build the starter — feel the client and operation model directly