You're running PostgreSQL + MongoDB + Redis + Pinecone + ClickHouse + TimescaleDB — six vendors, six SDKs, six dashboards, and $362–$4,900/month in combined bills. Magnus replaces all of them with one TypeScript SDK on Cloudflare's global edge.
Free tier · No credit card · Prisma-style TypeScript SDK
Magnus covers all six data models your application needs. Each model has first-class TypeScript types, auto-complete, and Prisma-quality DX — with smart routing to the right backend at the edge.
// Any SQL, full Postgres dialect const users = await db.sql` SELECT * FROM users WHERE plan = 'pro' AND created_at > ${cutoff} `
// MongoDB-compatible API const orders = await db .docs('orders') .find({ status: 'pending' }) // Simple ops → D1 edge (<5ms) // Complex aggs → MongoDB
// Redis-compatible + strong consistency await db.kv.set( 'session:' + token, sessionData, { ttl: 3600 } )
// Semantic search, embeddings const results = await db.vector.search({ query: userQuestion, collection: 'product-docs', topK: 10, rerank: true })
// High-cardinality telemetry await db.timeseries.write({ metric: 'api.latency', value: 8.3, tags: { region: 'eu-west' } })
// OLAP queries, aggregations const report = await db.analytics.query(` SELECT region, sum(revenue) AS total FROM events WHERE date > '2026-01-01' GROUP BY region `)
A typical production SaaS stack in 2026 costs $362–$4,900/month across six vendors. Magnus covers all six models starting at $99/month — with one invoice, one dashboard, and one SDK.
One import. Full TypeScript types. Auto-complete for SQL, document filters, vector queries, and time-series writes. Magnus infers which backend to use — you just write the query.
db.docs('orders').find<Order>() returns typed results@magnus/mcp into any agent framework for direct database accessimport { Magnus } from '@magnus/client' const db = new Magnus({ projectId: 'my-app', apiKey: process.env.MAGNUS_KEY }) // SQL — any Postgres dialect const users = await db.sql` SELECT * FROM users WHERE plan = 'pro' ` // Documents — MongoDB-compatible const orders = await db .docs('orders').find<Order>({ status: 'pending' }) // KV / Cache await db.kv.set('session:' + id, data, { ttl: 3600 }) // Vector search const similar = await db.vector.search({ query: userQuery, topK: 10 }) // All six models. One import. One bill.
Every tier includes all six data models. No per-model add-ons. No surprise bills at the end of the month.
All six models. Perfect for side projects.
For teams building production applications.
Replaces your entire $362–$4,900/mo stack.
Dedicated infrastructure for data-intensive products.
Compare: PostgreSQL + MongoDB + Redis + Pinecone + ClickHouse + TimescaleDB = $362–$4,900/mo
"We were paying $1,200/month across Supabase, MongoDB, and Pinecone. Magnus Pro is $99. Same capabilities, one connection string, TypeScript types for everything. This should not have been this easy."
"The SQL-to-MongoDB auto-routing is wild. I write a simple find() and it hits D1 at the edge in 4ms. I write an $aggregate with $lookup and it silently routes to MongoDB. I don't have to think about which database to use for which query."
"The MCP server is the unlock for us. Our agents can query SQL, search vectors, and cache results — all through @magnus/mcp in the agent config. Zero custom code between the agent and the database layer."
Magnus runs on Cloudflare's infrastructure: D1 (SQLite/Postgres-dialect) for relational and simple document ops, Vectorize for vectors, KV + Durable Objects for cache/KV, Analytics Engine for OLAP. Complex MongoDB aggregations can route to an optional MongoDB backend. One API, best-fit storage under the hood.
Simple document operations (find, insert, update, delete, basic queries) are fully compatible — served from D1 at the edge. Complex aggregations ($lookup, $facet, $text search) are supported via optional MongoDB right-behind routing. The SDK handles the routing transparently.
Yes. Magnus accepts standard Postgres DDL migrations and pg_dump output. Your existing ORM (Prisma, Drizzle, TypeORM) works without modification — just update your connection string. Data migration guides available for Supabase, Neon, and RDS.
Add @magnus/mcp to your agent's MCP config. Your agent gets tools for SQL queries, document operations, vector search, KV reads/writes, and time-series — no custom wrappers needed. Works with Claude, Cursor, LangGraph, and any MCP-compatible framework.
Reads from D1 at the nearest Cloudflare edge node: <5ms p95 for cached reads. Vector searches from Vectorize: <20ms for 1M vectors. KV reads from Cloudflare KV: <10ms globally. Write paths go through your primary region and are typically 20–50ms.
Start with what you need — all models are available on every tier, you just pay for what you use in terms of storage and operation counts. The SDK never forces you to use all six. And if you add a new model later, there's no migration — it's already in your existing Magnus project.
Free tier. All six models. Prisma-quality TypeScript SDK.
Hobby free · Builder $29/mo · Pro $99/mo · Scale custom