Hello from SecantusDB

Welcome to the SecantusDB blog. SecantusDB is a real MongoDB server written in Python, backed by the same WiredTiger engine MongoDB ships with. It speaks the MongoDB wire protocol on a real TCP socket, so pymongo, mongo-go-driver, mongosh, and mongodump all connect unchanged. Within single-node scope, your driver can't tell us apart from mongod.

What it's for

• Tests. Spin up an embedded server in a fixture, run your suite, tear it down. No Docker, no ports to manage, parallel-test friendly.
• Dev loops. Local development without standing up a mongod process. Default storage is on-disk WiredTiger, so your data survives a restart.
• Single-node prototypes. Small services that don't need a real cluster but want a real document database with real durability.

What it isn't

SecantusDB is not a replacement for production MongoDB. It does single-node only by design. Replica sets, sharding, real cluster topology, and anything that depends on multi-node behaviour are out of scope. We advertise ourselves as a single-node secantus replica-set primary in the hello reply so pymongo's topology machinery accepts change streams, but the topology is fictional — there is no other member.

Status: alpha

We're in early development. The Python API surface (CLI flags, public class signatures) is still settling and may shift between point releases. The on-disk format is WiredTiger's, and our schema on top of it (collection / index / oplog tables) has been stable across releases. There's no migration tool yet, so don't put production data here.

What's next

More driver-conformance coverage, performance work on the aggregation pipeline, and richer index acceleration (multi-field sort, collation). Follow along on GitHub or read the docs.