Connect
Copy the connection URL from the database’s page in the dashboard. It has this shape, with thedefault user authenticating against the admin database:
.pem under Connection → SSL/TLS certificate and pass it as both the client certificate and the CA: tls=true, tlsCertificateKeyFile and tlsCAFile.
- Node.js
- Mongoose
- Python
- mongosh
A standalone server
Each database is a single standalone server: no replica set and no oplog, and it can’t be converted into one. That means:
Single-document writes are atomic: updates with operators such as
$inc, $push or $set, and findOneAndUpdate, cover many cases that would otherwise need a transaction. Libraries that open sessions with transactions (for example Mongoose’s connection.transaction()) fail with the first error above.
Connections
Each database accepts up to 256 simultaneous connections. EveryMongoClient keeps its own pool (maxPoolSize) plus 2 monitoring connections, so create one client per process and reuse it: a client per request runs out of connections quickly. The Insights tab shows who holds the open connections.
Your user and permissions
Thedefault user has the readWriteAnyDatabase and dbAdminAnyDatabase roles. You can create and drop databases, collections and indexes, and read db.stats(), $collStats, explain() and the profiler (system.profile).
It can’t create users or roles, run db.serverStatus(), see other connections’ operations (currentOp), or run top, hostInfo, setParameter or shutdown. Monitoring tools that depend on serverStatus don’t work; use db.stats(), $collStats and the profiler instead.
Memory
MongoDB sizes its WiredTiger cache from the database’s RAM: the larger of 50% of (RAM minus 1 GB) and 256 MB.- The size is read when the server starts: restart the database after changing its RAM. It can’t be tuned.
- The cache holds only the data and index pages your queries touched.
- Memory outside the cache: about 1 MB per connection, index builds, and aggregations (a stage above 100 MB spills to disk).
- The memory allocator keeps freed memory for reuse, so the RAM shown in the dashboard climbs and then stays high. That is normal, not a leak.
Slow queries
The profiler records operations that take 100 ms or more in each database’ssystem.profile collection (capped at 1 MB), and the Insights tab lists them:
explain("executionStats") on a slow query to see whether it uses an index.

