Azure Cosmos DB Shell quick start

適用対象: ✅ NoSQL

Get started with Azure Cosmos DB Shell in just a few minutes with these practical examples.

Prerequisites

Azure Cosmos DB Shell installed (Installation Guide)
Azure Cosmos DB account
Authentication configured (Microsoft Entra ID, Managed Identity, or Account Keys)

Launch the shell

cosmosdbshell

You'll see a prompt:

CS >

Connect to your account

When launched without connection arguments, Azure Cosmos DB Shell starts in a disconnected state. Use the connect command to authenticate and connect to an account.

Use Microsoft Entra ID (recommended):

CS > connect https://<account-name>.documents.azure.com:443/ --auth-method entra-id

This starts the browser sign-in flow.

Use Managed Identity (production):

CS > connect https://<account-name>.documents.azure.com:443/ --auth-method managed-identity

Use Account Key (development/testing):

CS > connect DefaultEndpointProtocol=https;AccountName=<account-name>;AccountKey=<account-key>; --auth-method key

View your account endpoint

CS > endpoint

List databases

CS > ls

Output:

database1
database2
mydb

Navigate to a database

CS > cd mydb
CS >

List containers in database

CS > ls

Output:

users
products
orders

Navigate to a container

CS > cd users
CS >

Show current path

CS > pwd

Output:

/mydb/users

Basic operations

Create database

CS > mkdb mynewdb

Create container

CS > mkcon mycontainer -pk /id

(Creates container with partition key /id)

Query documents

CS > query "SELECT * FROM c WHERE c.status = 'active'"

Output:

{
  "id": "user1",
  "name": "Alice",
  "status": "active"
}
{
  "id": "user2", 
  "name": "Bob",
  "status": "active"
}

Count documents

CS > query "SELECT COUNT(*) FROM c"

Insert document

CS > create item {"id": "user3", "name": "Charlie", "status": "active"}

Update document

CS > update {"id": "user1", "name": "Alice", "status": "inactive"}

Delete document

CS > rm user1

Delete container

CS > rmcon users

Delete database

CS > rmdb mydb

Piping commands

Combine commands to create powerful workflows.

Query and count

CS > query "SELECT * FROM c" | jq 'length'

Filter results with jq

CS > query "SELECT * FROM c" | jq '.[] | select(.status == "active")'

Extract specific fields

CS > query "SELECT * FROM c" | jq '.[] | {id, name}'

Transform and output

CS > query "SELECT * FROM c" | jq '.[] | .name' | tr '\n' ','

Shell scripts

Create a script file for automated operations.

Example script: `batch_operations.sh`

#!/bin/bash

# Connect to Cosmos DB Shell
cosmosdbshell << 'EOF'
cd mydb
cd users

# Insert multiple documents
create item {"id": "user10", "name": "David", "status": "active"}
create item {"id": "user11", "name": "Eve", "status": "inactive"}
create item {"id": "user12", "name": "Frank", "status": "active"}

# Query and count
query "SELECT COUNT(*) as count FROM c"

# Exit
exit
EOF

Run script

bash batch_operations.sh

Filtering and searching

Find documents with specific criteria

CS > query "SELECT * FROM c WHERE c.age > 30 AND c.status = 'active'"

Search in arrays

CS > query "SELECT * FROM c WHERE ARRAY_CONTAINS(c.tags, 'vip')"

Aggregate results

CS > query "SELECT c.status, COUNT(*) as count FROM c GROUP BY c.status"

Export and import

Export data to JSON

CS > query "SELECT * FROM c" > users_export.json

The exported file wraps the documents in an items envelope:

{
  "items": [
    { "id": "user1", "pk": "tenantA", "name": "Alice", "status": "active",
      "_rid": "...", "_self": "...", "_etag": "...", "_attachments": "...", "_ts": 1730000000 },
    { "id": "user2", "pk": "tenantA", "name": "Bob", "status": "active",
      "_rid": "...", "_self": "...", "_etag": "...", "_attachments": "...", "_ts": 1730000001 }
  ]
}

Export to CSV-like format

CS > query "SELECT * FROM c" | jq -r '.id, .name, .status' | paste -sd, > users.csv

Import a bare JSON array

The create item command (alias mkitem) accepts either a single JSON object or a top-level JSON array. Use this form when you control the input file and can produce a clean array of documents.

Sample users_import.json:

[
  { "id": "user10", "pk": "tenantA", "name": "David", "status": "active" },
  { "id": "user11", "pk": "tenantA", "name": "Eve",   "status": "inactive" },
  { "id": "user12", "pk": "tenantB", "name": "Frank", "status": "active" }
]

Import it by piping the file into mkitem:

CS > cat users_import.json | mkitem

Add --force (alias --upsert) to replace existing items instead of failing on id conflicts:

CS > cat users_import.json | mkitem --force

Import a previously exported query result

A file produced by query "SELECT * FROM c" > users_export.json is not a bare array — it has the { "items": [ ... ] } wrapper shown above and the documents still contain Cosmos system fields (_rid, _self, _etag, _attachments, _ts) that must not be sent back on insert.

Use jq to unwrap the array and strip system fields before piping to mkitem:

CS > cat users_export.json \
  | jq '[.items[] | del(._rid, ._self, ._etag, ._attachments, ._ts)]' \
  | mkitem --force --db=mydb --con=users

What each step does:

.items[] unwraps the envelope and streams each document.
del(...) removes the read-only Cosmos system fields.
[ ... ] re-collects the results into the top-level array mkitem expects.
--force makes the import idempotent (re-running replaces existing items).
--db / --con target the destination explicitly so the import works regardless of your current shell location.

mkitem iterates the array sequentially and prints a summary with the number of items created or replaced and the total RU charge. Per-item failures (for example, an id conflict without --force) are reported inline but do not stop the import.

Tips and best practices

Use aliases for common commands

# Add to your shell profile
alias cosmosdb='cosmosdbshell'

Tab completion

Use Tab to auto-complete database and container names
Press Tab twice to see all available options

Command history

Arrow up/down to navigate command history
Ctrl+R to search command history

Formatting output

# Pretty-print JSON
CS > query "SELECT * FROM c" | jq .

# Compact output
CS > query "SELECT * FROM c" | jq -c .

Performance tips

Use partition key in queries for faster results:

CS > query "SELECT * FROM c WHERE c.id = 'user1'"

Limit result set for large queries:
```
CS > query "SELECT TOP 100 * FROM c"
```

Use projections to reduce data transfer:

CS > query "SELECT c.id, c.name FROM c"

Common workflows

Bulk insert from file

CS > jq -r '.[]' data.json | while read line; do
  create $line
done

Backup container

CS > query "SELECT * FROM c" > backup_users.json

Verify data migration

CS > query "SELECT COUNT(*) FROM c" 
CS > query "SELECT COUNT(*) FROM c"

Next steps

Complete Command Reference - Learn all available commands
MCP Server Setup - Enable AI integration
Troubleshooting Guide - Resolve issues
Security Best Practices - Secure your setup

フィードバック

このページはお役に立ちましたか?

Last updated on 2026-05-15

Azure Cosmos DB Shell quick start

Prerequisites

Launch the shell

Connect to your account

Basic navigation

View your account endpoint

List databases

Navigate to a database

List containers in database

Navigate to a container

Show current path

Basic operations

Create database

Create container

Query documents

Count documents

Insert document

Update document

Delete document

Delete container

Delete database

Piping commands

Query and count

Filter results with jq

Extract specific fields

Transform and output

Shell scripts

Example script: batch_operations.sh

Run script

Filtering and searching

Find documents with specific criteria

Search in arrays

Aggregate results

Export and import

Export data to JSON

Export to CSV-like format

Import a bare JSON array

Import a previously exported query result

Tips and best practices

Use aliases for common commands

Tab completion

Command history

Formatting output

Performance tips

Common workflows

Bulk insert from file

Backup container

Verify data migration

Next steps

See also

フィードバック

その他のリソース

Example script: `batch_operations.sh`