diff --git a/README.md b/README.md
index 081de5f..d1b3cee 100644
--- a/README.md
+++ b/README.md
@@ -1,32 +1,82 @@
 # Conduit Connector MongoDB
 
-## General
-
 The [MongoDB](https://www.mongodb.com/) connector is one of Conduit plugins. It
 provides both, a source and a destination MongoDB connector.
 
-### Prerequisites
+<!-- readmegen:description -->
+## Source
 
-- [Go](https://go.dev/) 1.23+
-- [MongoDB](https://www.mongodb.com/) [replica set](https://www.mongodb.com/docs/manual/replication/) (
-  at least single-node)
-  or [sharded cluster](https://www.mongodb.com/docs/manual/sharding/)
-  with [WiredTiger](https://www.mongodb.com/docs/manual/core/wiredtiger/)
-  storage engine
-- [Docker](https://www.docker.com/)
-- (optional) [golangci-lint](https://github.com/golangci/golangci-lint) v1.55.2
+The MongoDB Source Connector connects to a MongoDB with the provided `uri`, `db`
+and `collection` and starts creating records for each change detected in a
+collection.
 
-### How to build it
+Upon starting, the Source takes a snapshot of a given collection in the
+database, then switches into CDC mode. In CDC mode, the plugin reads events from
+a [Change Stream](https://www.mongodb.com/docs/manual/changeStreams/). In order
+for this to work correctly, your MongoDB instance must
+meet [the criteria](https://www.mongodb.com/docs/manual/changeStreams/#availability)
+specified on the official website.
 
-Run `make build`.
+### Snapshot Capture
 
-### Development
+When the connector first starts, snapshot mode is enabled. The connector reads
+all rows of a collection in batches using
+a [cursor-based](https://www.mongodb.com/docs/drivers/go/current/fundamentals/crud/read-operations/cursor/)
+pagination,
+limiting the rows by `batchSize`. The connector stores the last processed
+element value of an `orderingColumn` in a position, so the snapshot process can
+be paused and resumed without losing data. Once all rows in that initial
+snapshot are read the connector switches into CDC mode.
 
-Run `make install-tools` to install all the required tools.
+This behavior is enabled by default, but can be turned off by adding
+`"snapshot": false` to the Source configuration.
 
-Run `make test` to run all the units and `make test-integration` to run all the
-integration tests, which require Docker to be installed and running. The command
-will handle starting and stopping docker container for you.
+### Change Data Capture
+
+The connector implements CDC features for MongoDB by using a Change Stream that
+listens to changes in the configured collection. Every detected change is
+converted into a record and returned in the call to `Read`. If there is no
+available record when `Read` is called, the connector returns
+`sdk.ErrBackoffRetry` error.
+
+The connector stores a `resumeToken` of every Change Stream event in a position,
+so the CDC process is resumble.
+
+> **Warning**
+>
+> [Azure CosmosDB for MongoDB](https://learn.microsoft.com/en-us/azure/cosmos-db/mongodb/change-streams)
+> has very limited support for Change Streams, so they cannot be used for CDC.
+> If CDC is not possible, like in the case with CosmosDB, the connector only
+> supports detecting insert operations by polling for new documents.
+
+### Key handling
+
+The connector always uses the `_id` field as a key.
+
+If the `_id` field is `bson.ObjectID` the connector converts it to a string when
+transferring a record to a destination, otherwise, it leaves it unchanged.
+
+## Destination
+
+The MongoDB Destination takes a `opencdc.Record` and parses it into a valid
+MongoDB query. The Destination is designed to handle different payloads and
+keys. Because of this, each record is individually parsed and written.
+
+### Collection name
+
+If a record contains an `opencdc.collection` property in its metadata it will be
+written in that collection, otherwise it will fall back to use the `collection`
+configured in the connector. Thus, a Destination can support multiple
+collections in the same connector, as long as the user has proper access to
+those collections.
+
+### Key handling
+
+The connector uses all keys from an `opencdc.Record` when updating and deleting
+documents.
+
+If the `_id` field can be converted to a `bson.ObjectID`, the connector converts
+it, otherwise, it uses it as it is.<!-- /readmegen:description -->
 
 ### Source Configuration Parameters
 
@@ -248,4 +298,16 @@ pipelines:
 ```
 <!-- /readmegen:destination.parameters.yaml -->
 
+### How to build it
+
+Run `make build`.
+
+### Development
+
+Run `make install-tools` to install all the required tools.
+
+Run `make test` to run all the units and `make test-integration` to run all the
+integration tests, which require Docker to be installed and running. The command
+will handle starting and stopping docker container for you.
+
 ![scarf pixel](https://static.scarf.sh/a.png?x-pxid=528a9760-d573-4524-8f65-74a5e4d402e8)
diff --git a/source/iterator/cdc.go b/source/iterator/cdc.go
index 2c1a7eb..8aafe00 100644
--- a/source/iterator/cdc.go
+++ b/source/iterator/cdc.go
@@ -39,12 +39,12 @@ const (
 var changeStreamMatchPipeline = bson.D{
 	{
 		Key: "$match", Value: bson.M{
-		"operationType": bson.M{"$in": []string{
-			operationTypeInsert,
-			operationTypeUpdate,
-			operationTypeDelete,
-		}},
-	},
+			"operationType": bson.M{"$in": []string{
+				operationTypeInsert,
+				operationTypeUpdate,
+				operationTypeDelete,
+			}},
+		},
 	},
 }