MCP Server

The Adeptia Automate MCP Server is an AI-native access layer for Adeptia Automate. Built on the Model Context Protocol (MCP), it lets hosted AI assistants — such as ChatGPT and Claude.ai — connect directly to your Adeptia Automate instance and reason about it the way an integration engineer would. From a single connection, an AI assistant can list your workflows, inspect execution history, drill into step-level errors, and pull up dashboard-style summaries of what your integrations are doing.

	The MCP Server is the same backend that powers Observe. Both Observe and the MCP Server are packaged together in the adeptia-automate-mcp Helm chart. They are not deployed as part of the standard Adeptia Automate installation — you deploy them separately as an independent Helm release after the main application is running.

Connecting to the MCP Server

Hosted AI clients (ChatGPT, Claude.ai, and any other hosted assistant that supports remote MCP servers) connect over HTTP and authenticate using OAuth 2.1.

Point your AI client at your MCP Server's /mcp endpoint (for example, https://mcp.your-company.com/mcp). The client discovers the authorization server automatically and initiates an OAuth 2.1 sign-in flow — no client ID, client secret, or pre-registration is required.

The first time you connect, you'll be redirected to a login page where you enter:

• Instance URL — your Adeptia Automate URL
• Username — your Adeptia Automate username
• Password — your Adeptia Automate password

Whether the Instance URL is required during login depends on how the MCP server is deployed. In a single-instance deployment, where ADEPTIA_API_URL is pre-configured on the MCP server, the login page prompts only for a username and password — the Instance URL is not required as it is already set on the server. In a multi-instance or shared MCP deployment, where ADEPTIA_API_URL is not set, the login page will additionally prompt for the Instance URL along with the username and password.

Once you sign in, the assistant is connected to your instance and the Adeptia tools become available in the conversation. Your credentials are never stored by the MCP Server. Sessions last 24 hours, after which you'll be prompted to sign in again.

Configuring the mcp endpoint in AI Client 

The steps to configure mcp endpoint may vary depending on the AI client you use. The following steps illustrate how to configure the endpoint in Claude desktop:

1. Navigate to Settings > Developer.
2. Click Edit Config.
3. In the claude_desktop_config.json file, add MCP endpoint as shown below:

Code

{
  "mcpServers": {
    "adeptia": {
      "url": "<https://<host>/mcp>"
    }
  }
}

4. Restart Claude desktop.
5. Claude Desktop will open your browser where you need to enter your username and password for the Adeptia instance to authenticate.
6. Once authenticated, look for the hammer icon (🔨) in the Claude Desktop chat input bar. Click it to view the list of tools loaded from the Adeptia MCP server.

Using the MCP Server

Once connected, you can ask your AI assistant natural-language questions about your Adeptia Automate instance and have it use the MCP Server's tools on your behalf. You do not need to remember individual tool names — ask your question in plain language and the assistant will select the appropriate tools, call them with the right parameters, and return a structured response inline.

A typical investigation might combine several tools. For example, to investigate a failing integration, the assistant might first call list_workflows to find the workflow, then get_execution_history to see its recent runs, and finally get_execution_details to surface the step-level error.

Available tools

The MCP Server currently exposes the following monitoring tools. All tools are available in Private Preview.

Tool	Description
list_workflows	List all workflows with filtering by status, project, owner, and search.
get_workflow_status	Get detailed workflow metadata and current status by workflow ID.
get_execution_history	Get workflow execution history with status, timing, and filtering by automation and partner.
get_execution_details	Get execution logs with a step-level breakdown (Source, Adeptia, Destination), and optional audit logs and execution context.
get_automation_overview	Get a comprehensive automation dashboard: summary, trends, completed and pending automations, automation types, and process flows.
get_trigger_dashboard	Get the trigger and event dashboard: event chart, triggered processes, active triggers, process errors, and event errors.
list_transactions	List all automations with filtering and search.
get_transaction_details	Get detailed automation information by automation ID.
get_diagnostics	Get diagnostic information from the instance, including version, build tag, and application settings.

Example prompts

Once connected, you can ask your AI assistant questions like:

• "Show me all workflows in project X and their execution status over the last week."
• "Why did the most recent run of the EDI inbound workflow fail? Drill into the execution steps."
• "Give me an overview of automation activity across all projects today."
• "List the most recent automations and flag any that failed."
• "What is the current version of our Adeptia Automate instance?"

The assistant determines which tools to use, executes them with the correct inputs, and returns a structured response inline.