How to Import CSV to MongoDB (4 Methods)

MongoDB stores data as documents, not rows, so importing a CSV involves converting tabular data into a document structure. The tools handle this automatically -- each CSV row becomes a document, each column becomes a field. Here are the practical ways to do it.

Method 1: mongoimport (CLI)

mongoimport is part of the MongoDB Database Tools package. It's the fastest way to load CSV data from the command line:

mongoimport \
  --uri "mongodb://localhost:27017/mydb" \
  --collection users \
  --type csv \
  --headerline \
  --file users.csv

Key flags:

• --headerline -- uses the first row as field names
• --type csv -- specifies the file format (also supports tsv and json)
• --drop -- drops the collection before importing (useful for full refreshes)
• --ignoreBlanks -- skips fields with empty values instead of storing empty strings

Specifying field types. By default, mongoimport imports everything as strings. To specify types:

mongoimport \
  --uri "mongodb://localhost:27017/mydb" \
  --collection users \
  --type csv \
  --headerline \
  --columnsHaveTypes \
  --fields "name.string(),age.int32(),email.string(),signup_date.date(2006-01-02)" \
  --file users.csv

The --columnsHaveTypes flag with --fields lets you define the type for each column. Date parsing uses Go-style format strings (where 2006-01-02 represents YYYY-MM-DD).

Method 2: MongoDB Compass (GUI)

Compass is MongoDB's official GUI tool. It provides visual CSV import:

1. Connect to your MongoDB instance
2. Select the target database and collection (or create a new one)
3. Click Add Data > Import JSON or CSV file
4. Select your CSV file
5. Compass shows a preview with detected types -- adjust types per field if needed
6. Click Import

Compass lets you change individual field types (string, number, date, boolean) before importing, which is more user-friendly than mongoimport's type syntax.

Method 3: Python with pymongo

For programmatic control, use Python's csv module with pymongo:

import csv
from pymongo import MongoClient
from datetime import datetime
 
client = MongoClient('mongodb://localhost:27017/')
db = client['mydb']
collection = db['users']
 
documents = []
with open('users.csv', 'r') as f:
    reader = csv.DictReader(f)
    for row in reader:
        doc = {
            'name': row['name'],
            'age': int(row['age']) if row['age'] else None,
            'email': row['email'],
            'signup_date': datetime.strptime(row['signup_date'], '%Y-%m-%d')
        }
        documents.append(doc)
 
# Bulk insert for performance
collection.insert_many(documents)
print(f"Imported {len(documents)} documents")

For large files, batch the inserts to control memory usage:

BATCH_SIZE = 10000
batch = []
with open('users.csv', 'r') as f:
    reader = csv.DictReader(f)
    for row in reader:
        batch.append(transform(row))
        if len(batch) >= BATCH_SIZE:
            collection.insert_many(batch)
            batch = []
    if batch:
        collection.insert_many(batch)

Method 4: MongoDB Atlas UI

If you're using MongoDB Atlas (the managed cloud service):

1. Navigate to your cluster in the Atlas dashboard
2. Click Browse Collections
3. Select the target database and collection
4. Click Insert Document or use the Data Explorer import feature
5. Upload your CSV

Atlas also supports loading CSV data through Data Federation and scheduled triggers for recurring imports.

Common Gotchas

Type inference. mongoimport treats all values as strings unless you explicitly specify types with --columnsHaveTypes. This means "42" is stored as the string "42", not the number 42. If you plan to query by numeric range or sort numerically, you need explicit types.

Nested documents. CSV is flat -- there's no native way to represent nested objects. If you need nested structures (e.g., an address subdocument with city, state, zip), you'll need to transform the data during import using Python or a post-import aggregation pipeline.

Duplicate handling. By default, mongoimport stops on the first duplicate _id error. Use --mode=upsert to update existing documents or --mode=merge to merge fields:

mongoimport --uri "..." --collection users --type csv \
  --headerline --mode upsert --upsertFields email \
  --file users.csv

Date formats. The Go date format syntax used by mongoimport is unintuitive if you're used to strftime. The reference date is Mon Jan 2 15:04:05 MST 2006. So YYYY-MM-DD is 2006-01-02, and MM/DD/YYYY HH:MM:SS is 01/02/2006 15:04:05.

Large files. mongoimport handles large files well but uses a single thread. For multi-GB files, splitting the file and running parallel imports against the same collection is faster (MongoDB handles concurrent writes safely).

Mako connects to PostgreSQL, MySQL, MongoDB, BigQuery, Snowflake, and ClickHouse with AI-powered autocomplete. Try it free at mako.ai.