Learn how to add the Unstructured Transform MCP server to Dust as a remote MCP server. Your Dust agents can then point to your files and have Transform start producing partitioned, enriched, chunked, and embedded data based on your files in minutes.
Unlike local AI tools that run the Transform MCP server through a local bridge, Dust connects
directly to the hosted Transform MCP server as a remote MCP server. A Dust workspace admin adds the server
once by its URL; builders can then attach the Transform tools to any agent in the Agent Builder.
A Dust workspace, and the admin role within it. Only admins can add remote MCP servers. Sign in to Dust.
To use Bearer token authentication, you will need an Unstructured API key. This method of authentication is best for headless or shared-service setups. Get an API key.This is not required for you to use the recommended Automatic (OAuth) sign-in flow.
In the sidebar, go to Spaces, and then click Tools.
Click Add Tools, and then click Add MCP Server.
For the server URL, enter the following:
https://mcp.transform.unstructured.io
Choose an authentication method:
Automatic (recommended): Dust discovers the Transform MCP server’s OAuth configuration from the URL and
starts a sign-in flow. Follow the on-screen directions to sign in to Unstructured and approve access.
Bearer Token: Paste your Unstructured API key. Dust sends it as an Authorization header on every call to
the server. Use this for headless or shared-service setups.
Choose how credentials are shared:
Shared: one connection is used for the entire workspace.
Personal: each user authenticates individually the first time they use the tool.
Click Connect. Dust synchronizes the Transform tools and makes them available to your workspace.
The Transform MCP server is designed to notify Dust about these limits. Because of this, your Dust agent should
notify you whenever it encounters a file that exceeds 50 MB in size. Also, the agent should formulate strategies to
send requests that are 10 files or fewer and not cause more than 5 requests to be running at a time.
Start a conversation with an agent that has the Transform tools attached.
Provide the files you want Transform to partition, and ask the following:
Use the Unstructured Transform tools to partition these files.Provide the results as JSON, one JSON result per source file.
The agent drives the Transform job lifecycle for you: it requests an upload URL, uploads each file, starts the
transform, polls for status, and returns the results, one output per input file.
Sign-in fails before the Unstructured login screen appears. Your setup may not support automatic OAuth
discovery. Remove the server and re-add it using Bearer Token authentication with your Unstructured API key.
401 Unauthorized on tool calls. Re-check your API key (Bearer token), or re-run the OAuth flow (Automatic).
Members cannot add the server. Adding a remote MCP server is an admin-only action. Ask a workspace admin to
add it; members attach the synced tools to agents afterward.
Control Transform file partitioning output: Control how the Unstructured Transform MCP server instructs Transform to partition, enrich, chunk, and embed the data based on your files.
Control Transform generated sample code: Control how the Unstructured Transform MCP server generates sample curl or Python code that demonstrates how to use Transform to partition, enrich, chunk, and embed the data based on your files.