Integrate MongoDB with Langflow

You can integrate MongoDB with Langflow, an open-source visual framework for building AI pipelines and agents. Langflow provides a MongoDB Atlas component that uses MongoDB Vector Search as a vector store, so you can add MongoDB-backed retrieval to your flows without writing code.

The MongoDB Atlas component allows you to perform the following actions:

Ingest Data: Write documents and their embeddings to your collection.
Similarity Search: Run a similarity search and return matching documents.
Configure Agent Tools: Provide agents access to your MongoDB data.

Get Started

To complete a tutorial that configures Langflow to use MongoDB, see Get Started with the Langflow Integration.

Installation

The lfx-bundles package provides optional third-party integrations, including the MongoDB Atlas component. To install the MongoDB dependencies, run the following command:

uv pip install "lfx-bundles[mongodb]"

Alternatively, to install every bundled provider along with Langflow, run the following command:

uv pip install "langflow[bundles]"

Important

Docker Images

To use the MongoDB Atlas component with Docker, use the langflowai/langflow-all image instead of the the default langflowai/langflow image. The langflow-all image includes langflow[bundles], which provides the MongoDB Atlas component. To learn more, see Choose a Langflow Docker image.

Usage

To use MongoDB as a vector store in a flow, add the MongoDB Atlas component to your canvas from the MongoDB category, then configure its connection and index parameters in the component inspection panel.

The following tabs describe different ways to use the component in your flows:

To write documents and their embeddings to your collection, connect a data loader and a Split Text component to the Ingest Data input:

File or URL loader → Split Text → MongoDB Atlas

To run a similarity search and return matching documents, connect a Chat Input or other text-producing component to the Search Query input:

Chat Input → MongoDB Atlas → Chat Output

To give agents access to your collection, enable Tool Mode in the MongoDB Atlas component's header menu. Then, connect its output to an Agent component's Tools input:

MongoDB Atlas → Agent (Tools)

If you use the MongoDB Atlas component to query your collection, the component returns results that you can pass to other components in your flow as a list of JSON objects or as a tabular result. If both formats are available, you can set the format near the component's output port in the visual editor.

The first time you run a search, the component checks whether the MongoDB Vector Search index named in the Index Name parameter exists. If it doesn't, the component creates the index from the Index Field, Number of Dimensions, Similarity, and Quantization parameters, then waits 20 seconds before continuing. Because that wait doesn't confirm the index finished building, a query that runs immediately after index creation can still return empty results.

Tip

To check whether your index has finished building, view its Status in the Atlas UI.

Parameters

Some parameters are hidden by default in the visual editor. You can view and modify every parameter through the component inspection panel that appears when you select the component.

Parameter	Necessity	Description
MongoDB Atlas Cluster URI	Required	The connection string for your MongoDB Atlas cluster. To learn more about finding your connection string, see Connect to a Cluster via Client Libraries.
Enable mTLS	Optional	Whether to enable mutual TLS authentication. Defaults to `false`.
MongoDB Atlas Combined Client Certificate	Conditional	The client certificate combined with the private key. Required if you enable mTLS.
Database Name	Required	The name of the database that contains your collection.
Collection Name	Required	The name of the collection that stores your data.
Index Name	Required	The name of the MongoDB Vector Search index on your collection.
Ingest Data	Optional	The documents to write to the collection. Accepts one or more `Data`, `DataFrame`, or `Table` inputs from an upstream component.
Insert Mode	Optional	How the component inserts new documents into the collection. When set to `append`, the component keeps existing documents and adds the new documents. When set to `overwrite`, the component removes all existing documents from the collection before inserting the new documents. Defaults to `append`.
Embedding	Required	The embedding model that the component uses to embed your documents and queries. Connect an embedding model component to this input.
Search Query	Optional	The text query to run a similarity search with.
Number of Results	Optional	The number of results to return from a similarity search. Defaults to `4`.
Index Field	Optional	The field that the MongoDB Vector Search index covers. Defaults to `embedding`.
Filter Field	Optional	The field to filter the index on.
Number of Dimensions	Optional	The number of dimensions in the embedding vectors. Must match the number of dimensions that your configured embedder produces. Defaults to `1536`, but ensure that you set this value to match your embedding model's dimensions.
Similarity	Optional	The method used to measure similarity between vectors. The options are `cosine`, `euclidean`, or `dotProduct`. Defaults to `cosine`.
Quantization	Conditional	The quantization method that reduces memory costs by converting 32-bit floats to smaller data types. The options are `scalar` or `binary`, and there is no `none` option. Required if the MongoDB Vector Search index does not already exist. For more information, see the note in the Search Indexes section on this page.
Cache Vector Store	Optional	Whether to cache the vector store for the current component run. Defaults to `true`.

Search Indexes

The MongoDB Atlas component creates a MongoDB Vector Search index on your collection the first time you run a search, if a matching index doesn't already exist. You can also create or inspect this index manually. To learn more, see Create a MongoDB Vector Search Index.

The index uses the following definition, where the path field specifies your Index Field value, the numDimensions field specifies your Number of Dimensions value, and the similarity field specifies your Similarity value:

{
  "fields": [
    {
      "numDimensions": 1536,
      "path": "embedding",
      "quantization": "scalar",
      "similarity": "cosine",
      "type": "vector"
    }
 ]
}

If you set a Filter Field value, the component adds a filter type field for that path to the index definition.

Important

Required Values

Select a value for Quantization before you run a search. If you leave the field unset, index creation fails because the component sets the quantization field to null instead of omitting the field.

If your embedding model produces a different number of dimensions than the default 1536, ensure that you set the Number of Dimensions parameter accordingly.

Additional Resources

To learn more about using Langflow with MongoDB, see the following resources:

Langflow MongoDB Component documentation
Langflow GitHub repository

Back

Vertex AI Agent Engine

Get Started

Integrate MongoDB with Langflow

Get Started

Installation

Usage

Parameters

MongoDB Atlas Component ParametersUse these parameters to configure the MongoDB Atlas component.

Search Indexes

Additional Resources