Agentic Document Automation

Experlogix Smart Flows MCP server

Overview

Experlogix Smart Flows includes a Model Context Protocol (MCP) server. The Model Context Protocol is an open standard that describes how AI agents discover and call external tools. Because Smart Flows speaks this standard, any agent platform that supports MCP can work with Smart Flows in a structured and supported way.

Through the MCP server, an AI agent can discover the flows it is allowed to run, find out what input a flow expects, start an execution, follow its progress, and retrieve the documents that were produced. The agent does all of this on behalf of a Smart Flows user, and every execution stays inside the normal Smart Flows security and auditing framework.

This creates a bridge between conversational AI and governed document automation. A user has a conversation with an agent, and the agent decides when a document process is needed and which flow to run, while Smart Flows continues to enforce structure, permissions, and traceability behind the scenes.

What an Agent Can Do

Discover Flows

The agent asks the MCP server which flows are available and gets back the flows that the connected Smart Flows user is allowed to execute. This is normally the first thing an agent does, and it lets the agent answer questions such as which document processes can you run for me.

Understand What a Flow Needs

Before starting a flow, the agent retrieves the schema of that flow. The schema describes what the flow expects, for example a data payload, a record from a CRM or ERP system, or other structured input.

With this information the agent can collect the missing details from the user during the conversation, or validate the input it already has, instead of starting a flow that is bound to fail.

Start and Follow an Execution

The agent starts the execution and can then check its status, so it knows whether the flow is still running, has completed, or has failed.

When a flow needs the user to intervene, for example to fill in a form or to review a document, the MCP server returns a link to the Flow Execution Panel. The user opens that link, signs in, and continues the execution there. The agent therefore does not need to reproduce the flow interface in the conversation.

Retrieve the Generated Documents

After an execution has completed, the agent retrieves the list of documents that were produced, together with information about each of them, and can download the content. This allows the agent to present the result directly in the conversation, as a document or as a download link.

Typical Use Cases

Let an agent start document generation flows on behalf of a user.
Offer a conversational interface for flows that require structured input, where the agent collects the input first.
Return generated documents to the user through the agent, including direct download links.
Follow the progress and the outcome of a flow execution from within the agent platform.
Combine AI-driven interaction with automation logic that stays governed and reusable.

Tools Exposed by the MCP Server

Tool	Description
`get_flows`	Gets the list of flows that are available to execute.
`get_flow_schema`	Gets the JSON schema for a specific flow, which describes the input the flow expects.
`execute_flow`	Starts an execution of a specific flow.
`get_execution_status`	Gets the current status of a flow execution.
`get_documents`	Gets the list of documents generated by a completed flow execution.
`download_document`	Downloads the content of a specific document generated by a flow execution.
`sync_external_users`	Synchronizes external users from Smart Flows with all configured external connectors. This tool does not take any arguments.

Preparing Smart Flows

An agent authenticates against the MCP server with the API key of a Smart Flows user. Create a dedicated user for this purpose, give it exactly the permissions that the agent needs, and generate an API key for it.

Log in to the Smart Flows Project Console.
Open Control panel and select Users.
Open the user that the agent will act as, or create a new one.
Open the API keys tab.
Select Generate API key.
Give the key a meaningful name and confirm.
Copy the key from the Your new API key window.

Important: The key is shown only once. The window states: “This window will only be shown once, make sure you save the key”. Store the key as you would store a password. If it is ever compromised, delete it in the Project Console and generate a new one.

You also need the project URL of your Smart Flows project. Together with the API key, it identifies the project the agent connects to.

Connecting an Agent to Smart Flows

Any agent platform that supports the Model Context Protocol can connect to Smart Flows. There are two ways to register the connection, and which one applies depends on the platform:

The platform offers the Smart Flows MCP server in its own catalog, and you select it from a list.
The platform lets you register an MCP server yourself, and you provide its address and the authentication details.

Either way, the agent uses the same MCP server and the same set of tools. The sections below show the settings for two of the most widely used agent platforms:

Connecting from Microsoft Copilot Studio

The Smart Flows MCP server is published and discoverable in Microsoft Copilot Studio, so it can be added to an agent without any custom development.

In Copilot Studio, create an agent, or open an existing one.
Go to the Tools section of the agent and add a tool.
Select Model Context Protocol.
Select Experlogix Smart Flows MCP.
Create a new connection and provide:
- a display name for the connection (optional),
- the project URL of your Smart Flows project,
- the API key you generated at the Preparing Smart Flows step .
Confirm to add the tool and open its configuration.

The agent now has access to all tools of the MCP server. You can verify this in the tool overview, which lists the available operations.

Testing the Agent

Test the agent by asking it something that requires the MCP server, for example to list the flows it has access to.

The first time a tool is called, Copilot Studio may ask you to establish the connection. Open the connection manager, connect, and ask again. The agent then returns the list of flows that the connected Smart Flows user is allowed to execute.

Asking the agent to run one of those flows shows the full sequence: the agent locates the flow, retrieves its schema, and starts the execution. It returns a link to the Flow Execution Panel, where you sign in and continue the execution.

Registering the Smart Flows MCP Server in Agentforce AI

To register the Smart Flows MCP server as a connection in your Salesforce organization, use the following values:

Field	Value
MCP Server Name	A name of your choice, for example `ExperlogixDocuments`.
Server URL	`<https://mcp.smartflows.experlogix.com/runtime/webhooks/mcp`>
Authentication Method	OAuth 2.0
Header name	`X-API-Key`
Header value	`UserAPIKey;DAProjectURL`

The header value is a single string that combines the API key of a Smart Flows user and the URL of the Smart Flows project, separated by a semicolon (;).

Security and Governance

Every execution that is started through the MCP server runs inside the existing Smart Flows security, permission, and auditing framework. An agent is subject to exactly the same rules as any other execution channel.

The following applies:

The agent authenticates as a Smart Flows user, so it can only discover and execute the flows that this user is allowed to run.
Restricting what an agent can do is done by restricting the permissions of that user, not by configuring the agent.
Executions started by an agent are visible in monitoring and reporting, in the same way as executions started by a person.
An agent does not run under the identity of a user of the connected system. In Salesforce, for example, Salesforce permission sets do not apply to it.

Because of this, treat the integration user as a real account: give it the smallest set of permissions that is sufficient, and review that set when the responsibilities of the agent change.