11ai MongoDB CRUD

Version baseline: MongoDB 8.x, with 8.3.7 as the current stable release at this review. Inspect server, Atlas release track, featureCompatibilityVersion, mongosh, Database Tools, and driver versions independently before using 8.3-only behavior.

Work only after the deployment, database, collection, and document shape are known. Reads can run immediately when authorized; writes require a clear target, and multi-document changes require a count or preview before execution.

Create

db.users.insertOne({ email: "ada@example.com", active: true })
db.users.insertMany([
  { email: "a@example.com", active: true },
  { email: "b@example.com", active: true }
])

Validate required fields, types, uniqueness expectations, and whether the user wants generated _id values or supplied identifiers. Report inserted ids without exposing unrelated documents.

Read

Prefer the smallest useful projection and limit:

db.users.find(
  { active: true },
  { _id: 1, email: 1 }
).sort({ email: 1 }).limit(20)

Use findOne for a single expected document. Use countDocuments(filter) before a broad read or before any multi-document write. Do not dump an entire collection unless the user explicitly requests it and the data is safe to display.

Update

db.users.updateOne(
  { email: "ada@example.com" },
  { $set: { active: false }, $currentDate: { updatedAt: true } }
)

db.users.updateMany(
  { active: false, archived: { $ne: true } },
  { $set: { archived: true }, $currentDate: { updatedAt: true } }
)

Show the exact filter, update document, matched count, and modified count. Ask before running updateMany. Treat upsert as a write that can create a document; require a deliberate unique filter and explain the insert path.

Delete

db.users.deleteOne({ email: "ada@example.com" })
db.users.deleteMany({ archived: true })

For deleteOne, verify the filter identifies the intended document. For deleteMany, run countDocuments(filter), show the result, and request explicit approval for the exact filter. Never use an empty filter for deletion.

Verification

After a write, verify the result with a targeted read or result document. If the operation returns zero matches, do not broaden the filter automatically; inspect field names, types, database, and collection first.