microsoft/migrating-documentdb-to-cosmos
> Migrates from the deprecated Microsoft.Azure.DocumentDB SDK (V2) to the modern Microsoft.Azure.Cosmos SDK (V3) for Azure Cosmos DB. Use ONLY when Microsoft.Azure.DocumentDB has been flagged as deprecated or obsolete and must be replaced — not for version-bump scenarios where the V2 SDK is still supported.
npx skills add https://github.com/microsoft/upgrade-agent-plugins --skill migrating-documentdb-to-cosmos
Migrate from the legacy Microsoft.Azure.DocumentDB (V2) SDK to the modern Microsoft.Azure.Cosmos (V3) SDK. The V3 SDK introduces a new object model (CosmosClient → Database → Container) and requires explicit partition keys on most operations. All data-plane calls are async-first, and the Document base class is replaced by generic types or dynamic.
<PackageReference Include="Microsoft.Azure.DocumentDB" Version="{any}" />
<!-- Or the .NET Core variant -->
<PackageReference Include="Microsoft.Azure.DocumentDB.Core" Version="{any}" />
<PackageReference Include="Microsoft.Azure.Cosmos" Version="{latest-stable}" />
Use tools or NuGet to find the latest stable version.
Migration Progress:
- [ ] Step 1: Detect DocumentDB SDK usage
- [ ] Step 2: Update package references
- [ ] Step 3: Replace client and resource model
- [ ] Step 4: Convert CRUD operations
- [ ] Step 5: Update queries
- [ ] Step 6: Migrate error handling
- [ ] Step 7: Build and verify
Scan the project for:
using Microsoft.Azure.Documents; and using Microsoft.Azure.Documents.Client;DocumentClient, Document, DocumentCollection, ResourceResponseCreateDocumentAsync, ReadDocumentAsync, ReplaceDocumentAsync, DeleteDocumentAsyncRequestOptions and FeedOptions usageRemove Microsoft.Azure.DocumentDB or Microsoft.Azure.DocumentDB.Core from the project file. Add Microsoft.Azure.Cosmos.
// Old
var client = new DocumentClient(new Uri(endpoint), authKey);
// New
var client = new CosmosClient(endpoint, authKey);
var database = client.GetDatabase(databaseName);
var container = database.GetContainer(containerName);
The V3 SDK uses a hierarchy: CosmosClient → Database → Container. There is no need to construct URI links manually — use the fluent getter methods instead.
// Old — Create
ResourceResponse<Document> response = await client.CreateDocumentAsync(
collectionUri, document, new RequestOptions { PartitionKey = new PartitionKey(pk) });
// New — Create
ItemResponse<MyItem> response = await container.CreateItemAsync(
item, new PartitionKey(pk), new ItemRequestOptions());
// Old — Read
Document doc = await client.ReadDocumentAsync<Document>(
documentUri, new RequestOptions { PartitionKey = new PartitionKey(pk) });
// New — Read
ItemResponse<MyItem> response = await container.ReadItemAsync<MyItem>(
id, new PartitionKey(pk));
// Old — Delete
await client.DeleteDocumentAsync(documentUri,
new RequestOptions { PartitionKey = new PartitionKey(pk) });
// New — Delete
await container.DeleteItemAsync<MyItem>(id, new PartitionKey(pk));
// Old
var query = client.CreateDocumentQuery<MyItem>(collectionUri,
new SqlQuerySpec("SELECT * FROM c WHERE c.status = @status",
new SqlParameterCollection { new SqlParameter("@status", "active") }),
new FeedOptions { EnableCrossPartitionQuery = true });
// New
var query = container.GetItemQueryIterator<MyItem>(
new QueryDefinition("SELECT * FROM c WHERE c.status = @status")
.WithParameter("@status", "active"));
var results = new List<MyItem>();
while (query.HasMoreResults)
{
FeedResponse<MyItem> page = await query.ReadNextAsync();
results.AddRange(page);
}
Cross-partition queries are enabled by default in V3 — remove EnableCrossPartitionQuery settings.
// Old
try { /* operation */ }
catch (DocumentClientException ex) when (ex.StatusCode == HttpStatusCode.NotFound)
{ /* handle */ }
// New
try { /* operation */ }
catch (CosmosException ex) when (ex.StatusCode == HttpStatusCode.NotFound)
{ /* handle */ }
dotnet build
| DocumentDB V2 (Old) | Cosmos V3 (New) | Notes |
|----------------------|-----------------|-------|
| DocumentClient | CosmosClient | New connection model with builder options |
| Database (V2 resource) | Database (V3 proxy) | Obtained via CosmosClient.GetDatabase() |
| DocumentCollection | Container | Obtained via Database.GetContainer() |
| Document class | Generic T or dynamic | No base class required for items |
| CreateDocumentAsync | Container.CreateItemAsync<T> | Partition key is a required parameter |
| ReadDocumentAsync | Container.ReadItemAsync<T> | Takes id + partition key, not a URI |
| ReplaceDocumentAsync | Container.ReplaceItemAsync<T> | Takes id + partition key |
| DeleteDocumentAsync | Container.DeleteItemAsync<T> | Takes id + partition key |
| CreateDocumentQuery | Container.GetItemQueryIterator<T> | Uses QueryDefinition for parameterized queries |
| RequestOptions | ItemRequestOptions / QueryRequestOptions | Split into operation-specific options |
| FeedOptions | QueryRequestOptions | EnableCrossPartitionQuery removed (always on) |
| DocumentClientException | CosmosException | Unified exception type with StatusCode |
| URI-based resource links | Id-based fluent accessors | No manual URI construction needed |
The V3 SDK requires an explicit PartitionKey on most item operations. If the V2 code relied on a default or single-partition collection, add the partition key path to the container configuration and pass it with each call.
Document Base ClassV3 does not require items to inherit from Document. Replace Document with a plain POCO or dynamic. If code accesses Document.Id or Document.SelfLink, map Id to a property on the POCO and remove SelfLink references (V3 does not use self-links).
V3 queries return a FeedIterator that must be iterated with HasMoreResults / ReadNextAsync. Replace any synchronous AsEnumerable() or ToList() calls on the old IDocumentQuery with the async iteration pattern.
V3 accepts either an endpoint+key pair or an AccountEndpoint=...;AccountKey=... connection string. If the V2 code parsed connection strings manually, switch to the single-string CosmosClient(connectionString) constructor.
Take microsoft/migrating-documentdb-to-cosmos from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.