Why Muhkoo

Muhkoo Connect is a TypeScript SDK for building apps where the server is never trusted with user data. Authentication, storage, and messaging are all private by construction — yet you use them through one ergonomic client, the same way you’d use any BaaS.

The whole API is one object

const client = new Client({ apiKey, baseUrl });

client.auth      // ZK identity: register / login / sessions
client.kv        // per-user, encrypted-at-rest key/value
client.storage   // file storage (chunked, encrypted, erasure-coded)
client.message   // pub/sub + end-to-end-encrypted direct messages

Everything hangs off client.<namespace>.<method>. There’s no separate auth SDK, storage SDK, and realtime SDK to wire together — one session flows through all three.

What makes it different

• Zero-knowledge auth. A user’s identity is derived from their (username, password) on the device and proven with a Groth16 proof. The accelerator stores only a Poseidon commitment — never a password, never a secret. The same credentials reproduce the identity on any device, so login is federated with no key material to sync.
• Encryption by default. client.kv values are sealed with AES-256-GCM under a key derived from the user’s identity before they ever leave the device. Direct messages use the Double Ratchet (the Signal protocol’s ratchet). The platform relays ciphertext.
• Edge-native. The backend runs on the edge, so state lives close to users globally and there’s nothing to operate.
• Metered + multi-tenant. Every app authenticates with an app key (mk_…) that attributes usage for billing. The API is public — your users’ apps can run on any origin.

When to use it

Reach for Muhkoo Connect when you’re building:

• private, user-owned data apps (notes, journals, health, finance);
• end-to-end-encrypted chat or collaboration;
• multi-device apps that need federated login without a key-escrow story;
• anything where “the vendor can read your users’ data” is a non-starter.

When not to

• You need full-text search or server-side queries over user data — encrypted storage can’t be queried on the server by design (see Storage).
• You need the server to process plaintext (e.g. server-side AI over user content) — that breaks the trust model.

Ready? The Quickstart gets you from zero to a signed-in user reading and writing encrypted data in a few minutes.