Skip to content
Report an issue

Quickstart Guide: UiPath Coded MCP Servers

Introduction

This guide provides instructions for setting up and running a UiPath coded MCP Server.

Prerequisites

You need:

  • Python 3.11 or higher
  • pip or uv package manager
  • A UiPath Automation Cloud account with appropriate permissions
  • A UiPath Personal Access Token with Orchestrator API Access scopes

Creating a New Project

Use uv for package management. To create a new project:

mkdir examplecd example
New-Item -ItemType Directory -Path exampleSet-Location example
# Initialize a new uv project in the current directoryuv init . --python 3.11# Create a new virtual environment# By default, uv creates a virtual environment in a directory called .venvuv venvUsing CPython 3.11.16 interpreter at: [PATH]
Creating virtual environment at: .venv
Activate with: source .venv/bin/activate
# Activate the virtual environment# For Windows PowerShell/ Windows CMD: .venv\Scripts\activate# For Windows Bash: source .venv/Scripts/activatesource .venv/bin/activate# Install the uipath packageuv add uipath-mcp
# Create a new virtual environmentpython -m venv .venv# Activate the virtual environment# For Windows PowerShell: .venv\Scripts\Activate.ps1# For Windows Bash: source .venv/Scripts/activatesource .venv/bin/activate# Upgrade pip to the latest versionpython -m pip install --upgrade pip# Install the uipath packagepip install uipath-mcp

Create Your First UiPath Coded MCP Server

Create your first MCP server

uipath new math-serverβ ‹ Creating new mcp server 'math-server' in current directory ...
βœ“ Created 'server.py' file.
βœ“ Created 'mcp.json' file.
βœ“ Created 'pyproject.toml' file.
πŸ’‘ Initialize project: uipath init
════════════════════════════════════════════════════════════
Start 'math-server' as a self-hosted MCP server
════════════════════════════════════════════════════════════
πŸ’‘ 1. Set UIPATH_FOLDER_PATH environment variable
πŸ’‘ 2. Start the server locally: uipath run math-server

Warning

uipath new command deletes all previous .py files in the current directory.

This command creates the following files:

File Name Description
server.py A sample MCP math server using FastMCP
mcp.json Configuration file needed for coded UiPath MCP Servers.
pyproject.toml Project metadata and dependencies as per PEP 518.

Initialize Project

uipath initβ ‹ Initializing UiPath project ...
βœ“ Created '.env' file.
βœ“ Created 'uipath.json' file.

This command creates the following files:

File Name Description
.env Environment variables and secrets (this file is not packed & published).
uipath.json Input/output JSON schemas and bindings.

Authenticate With UiPath

uipath authβ ‹ Authenticating with UiPath ...
πŸ”— If a browser window did not open, please open the following URL in your browser: [LINK]
πŸ‘‡ Select tenant:
0: Tenant1
1: Tenant2
Select tenant number: 0
Selected tenant: Tenant1
βœ“ Authentication successful.

Run the MCP Server

There are two ways to run your coded MCP server:

1. Running Locally (On-Prem)

When running the server locally, JSON-RPC requests are tunneled from UiPath servers to your local server. During startup, the local server automatically registers itself with UiPath.

Since MCP servers are folder-scoped in Orchestrator, you need to set the UIPATH_FOLDER_PATH environment variable. To do this:

  1. Copy the folder path from your Orchestrator interface

  1. Add it to the .env file (created during uipath init) as: 
    UIPATH_FOLDER_PATH=<Copied folder path>
# Start the serveruipath run math-serverInitializing tracer instance. This should only be done once per process.
HTTP Request: GET https://***/orchestrator_/api/FoldersNavigation/GetFoldersForCurrentUser?searchText=Agents&skip=0&take=20 "HTTP/1.1 200 OK"
Folder key: ***
Initializing client session...
Initialization successful
Registering server runtime ...
...

Verifying the Server

Once started successfully, your MCP server will appear in Orchestrator. Navigate to the MCP Servers tab in your configured folder:

You can inspect the available tools by clicking on the server:

Now we can connect to the server using any MCP client. See the Connecting to the MCP Server section.

2. Running on UiPath Automation Cloud

Info

This quickstart guide provides instructions for deploying the MCP Server in My Workspace folder. Choosing this folder simplifies the configuration process, as you won’t need to manually handle the following:

  • Serverless machine allocation

  • Unattended robot permissions

  • Process creation (processes are automatically provisioned when a package is published to My Workspace)

If you prefer to deploy the MCP Server in a different folder, additional steps are required:

  1. Create a process from the MCP Server package.

  2. Ensure a serverless runtime (machine) is assigned to the target folder in Orchestrator.

  3. Confirm that a user with unattended robot permissions is assigned to the target folder.

To deploy your MCP server to UiPath Automation Cloud, follow these steps:

(Optional) Customize the Package

Update author details in pyproject.toml:

authors = [{ name = "Your Name", email = "your.name@example.com" }]

Package Your Project

uipath packβ Ή Packaging project ...
Name : math-server
Version : 0.0.1
Description: Description for math-server project
Authors : John Doe
βœ“ Project successfully packaged.

Publish The MCP Server Package

uipath publish --my-workspaceβ ™ Publishing most recent package: math-server.0.0.1.nupkg ...
βœ“ Package published successfully!
β ¦ Getting process information ...
πŸ”— Process configuration link: [LINK]
πŸ’‘ Use the link above to configure any environment variables

After publishing, you can configure and manage your MCP server through the UiPath Automation Cloud interface:

Configure in UiPath Automation Cloud

  1. In My Workspace, navigate to the MCP Servers tab and click Add MCP Server

  1. In the configuration dialog:
  • Select Coded as the server type
  • Choose the math-server process
  • Click Add to deploy the server

Once deployed, the server automatically starts and registers its available tools. You can monitor the job status in the MCP Server side panel.

Connecting to the MCP Server

You can connect to your MCP server using any MCP client. Here's what you need:

  1. MCP Server URL: Copy this from the UiPath MCP Servers page in Orchestrator

  1. Authentication: Use your Personal Access Token (PAT) with Orchestrator API Access scopes as authorization header
  2. Transport: Configure the client to use HTTP Streamable transport

Next Steps

Congratulations! You have successfully set up, created, published, and run a coded UiPath MCP Server. πŸš€

For more coded MCP samples, please refer to our samples section in GitHub.