Documentation Index

Fetch the complete documentation index at: https://docs.vention.com/llms.txt

Use this file to discover all available pages before exploring further.

Get Started: MachineAgent

Prev

Overview

MachineAgent brings agentic AI to the Vention platform. Using plain-language prompts, MachineAgent helps you lay out an automation cell, generate the industrial program that runs it, and analyze how deployed machines are performing — across the same design, program, and operate workflow you already use in MachineBuilder, MachineLogic, and MachineAnalytics.

MachineAgent is not a separate application. It is a set of agentic capabilities available at each stage of the automation lifecycle, plus open interfaces — the Vention CLI, the Vention MCP server, and Vention skills — that let you drive the platform from your own AI coding agent.

Requirements

  • A Vention account with access to MachineBuilder
  • For programming: a design containing a MachineMotion AI controller and at least one supported automation component (see Get started with MachineLogic)
  • For deployment to hardware: a MachineMotion AI controller connected to MachineCloud
  • For the CLI: a local development environment (Node.js) and an API token — see Command Line Interface (CLI)
  • For the MCP server: an AI agent that supports the Model Context Protocol, such as Claude Code, Claude Desktop, Cursor, VS Code, or OpenAI Codex

What MachineAgent Does at Each Stage

Stage Capability Where you use it
Design Generate and edit cell layouts from a description Your Agent + MCP
Program Generate, modify, and debug applications from a prompt MachineLogic Copilot
Program Drive the platform from your own AI coding agent Your Agent + Vention CLI + Developer Toolkit
Operate Analyze fleet performance, diagnose issues, and guide remediation Your Agent + MCP (or logs)

Installation and Setup

Connecting to the Vention MCP Server with Your AI Agent

The Vention MCP server exposes the Vention platform to any AI agent that supports the Model Context Protocol — including Claude Code, Claude Desktop, Cursor, VS Code, ChatGPT, and Gemini CLI. Once connected, your agent can work with your designs and query your deployed machines in the same conversation where you write your application code.

Connection Information

Endpoint https://mcp.vention.com/mcp
Transport Streamable HTTP
Authentication OAuth — no API token to create or store
Requires A Vention account

Add the endpoint URL to your agent's MCP server configuration. The first time the agent connects, your browser opens a Vention sign-in page. Approve the request and your agent is connected — there is no token to copy or paste.

Every MCP client is configured differently. Use your client's own documentation:

Client Setup guide
Claude Code Connect Claude Code to tools via MCPclaude mcp add --transport http vention https://mcp.vention.com/mcp
Claude Desktop and Cowork Get started with custom connectors using remote MCP
Cursor Model Context Protocol
VS Code (Copilot agent mode) MCP servers in VS Code
ChatGPT Developer mode and MCP apps in ChatGPT
Gemini CLI MCP servers
JetBrains AI Assistant Model Context Protocol

For any client not listed, see the MCP project's Connect to remote MCP servers guide and its directory of MCP clients.

Note: If your client can only launch local MCP servers, check for an update — every client above connects over HTTP directly. Local bridge tools such as mcp-remote are no longer needed and are not supported by Vention.

The server accepts up to 60 requests per minute per user, with a 256 KB request body limit.

Available Tools in the Vention MCP Server

These tools are available today:

Tool What it does
whoami Confirms which Vention account your agent is authenticated as
machine_status Returns the latest state reported by a MachineMotion controller, by serial number
machine_history State changes, run time, cycle data, and counter deltas over a time window
machine_health Machine vitals over time: CPU load and temperature, memory and disk usage, per-drive temperature
machine_activity_log A readable event feed including activity like: brake released, drive ready, state changed
machine_metric The raw series for a single numeric metric key, including custom counters from your application
machine_logs_request Requests a machine's logs be packaged for download
machine_logs_ready Returns a download link once the requested logs are packaged
design_versions_list Lists the versions of a design, with their floor plans
design_graph_download_url Creates a short-lived download link for a design version's assembly graph
design_graph_upload_create Creates a short-lived upload URL for a design's assembly graph
design_version_save Saves an uploaded graph as a new version of a design
design_floor_plan_create Creates a floor plan on a design
design_floor_plan_upload_create Creates a short-lived upload URL for a floor plan image

MachineAgent Skills

A skill is a markdown file that teaches your agent how Vention works — our conventions, workflows, and constraints. Skills complement the MCP server: the MCP server gives your agent access to the platform, while skills give it the judgment to use that access well.

The following skills are available:

Skill Use it for
vention-design Building and editing Vention design graphs — agentic machine layout
vention-programming Writing, debugging, and deploying MachineLogic applications
vention-monitoring Diagnosing deployed machines from controller state and analytics

To get the skill files, contact your Vention representative.

Installing Skills

Claude Code, Claude Desktop, and Cowork

Save the file as SKILL.md inside a folder named after the skill:

  • Available in every project: ~/.claude/skills/<skill-name>/SKILL.md
  • Available in one project only: .claude/skills/<skill-name>/SKILL.md in the project root

For example, to install the design skill for all your projects, save it as ~/.claude/skills/vention-design/SKILL.md. Restart your agent, and it will load the skill automatically when a task matches the skill's description.

Other agents

Cursor, VS Code, and other agents use their own filenames and locations for the same idea — Cursor calls them rules, VS Code calls them instructions. The body of each Vention skill works unchanged in any of them; you only need to place it where your client expects, and you can drop the --- frontmatter block at the top, which is Claude-specific.

Using Skills

Once installed, you do not invoke a skill directly — describe your task and your agent loads the matching skill on its own. For example:

  • "Lay out a Vention palletizing cell for 400 × 300 mm cases with a UR10e" loads the design graph skill
  • "Write a Vention MachineLogic app that indexes the conveyor and picks with the gripper" loads the programming skill
  • "Machine 6060009 keeps faulting, what's going on?" loads the monitoring skill

Note: Skills are guidance, not guardrails. Your agent can still be wrong. Review generated designs and applications before you build or deploy them, and treat any diagnosis as a hypothesis to confirm rather than a conclusion.

Using MachineAgent

1. Designing a Facility Layout

MachineAgent can produce a starting cell layout in MachineBuilder from a plain-language description of the application — for example, the part being handled, the process, the robot, and the footprint available.

Use the Vention design skill and Vention MCP Server.

Note: Generated layouts are a starting point, not a validated machine. Review them with the Automated Design Checker and have any safety-relevant design reviewed by a qualified engineer before building.

2. Programming with MachineLogic Copilot

MachineLogic Copilot is the agentic programming interface inside MachineLogic. It has context of your machine's Configuration and any robot Scene Assets already in your design, so you can refer to your actuators, robots, and outputs by name directly in a prompt.

There are two ways to use it:

Generate a new application from a prompt

  1. Open the MachineLogic tab in MachineBuilder.
  2. Choose Generate app with Vention Copilot.
  3. Describe what you want the application to do. Copilot creates the application and opens it for you.

Modify or debug an existing application

  1. Open a Python application in MachineLogic's Python editor.
  2. Open the Copilot tab in the bottom panel (use the Vention Extension toggle in the left side panel if the bottom panel is hidden).
  3. Enter a prompt. Copilot proposes code changes that you Apply and Accept directly in the application.

Tip: Paste errors from the Logs tab straight into Copilot to diagnose and fix a failing application quickly.

Copilot writes against the machine-logic-sdk Python API. For the full editor reference, see the MachineLogic Python Programming Guide.

3. Programming with the Vention CLI and an External AI Agent

The Vention CLI lets you work on MachineLogic applications on your own machine, in your own editor, with your own AI coding agent — including Claude Code. This is the recommended path for complex applications, version control, and teams that already have an agentic development workflow.

The workflow is:

  1. Install and authenticate the Vention CLI.
  2. Pull the application and its machine configuration to your local workspace.
  3. Work locally with your agent of choice against the Developer Toolkit and machine-logic-sdk.
  4. Push the application back to the platform, then simulate or deploy it.

See Command Line Interface (CLI) and Introduction to the Vention Developer Toolkit.

4. Monitoring and Troubleshooting Deployed Machines

MachineAgent can analyze equipment performance data and logs across a fleet of deployed machines — identifying an underperforming system, diagnosing the operational issue, and guiding you through remediation in plain language.

Use the Vention monitoring skill and Vention MCP Server.

This builds on the data your machines already report to MachineAnalytics:

Related Articles