◫OpenStudio AI
OpenStudio + AI agents

Agentic energy modeling, ready for the way you work.

OpenStudio AI brings guided model editing, HVAC creation, simulation, and results workflows to Codex and Claude Code. OpenStudio-MCP adds specialized local execution when it is part of your workflow.

Install for your client →Explore OpenStudio-MCP ↓
Codex + Claude CodeAgent-facing modeling workflows
20+ skillsReusable modeling workflows
DOE multi-lab effortMore MCPs to come
A translucent building energy model with highlighted HVAC paths and simulation data
A guided workflow, not a black box

A single, agentic front end for building-energy modeling.

Modelers work through the AI host they already use. OpenStudio AI coordinates the workflow, preserves the engineering record, and connects the right local execution path when it is available.

⌘

Edit models safely

Inspect and edit OpenStudio models with reviewed SDK patterns and a durable workflow state.

⌁

Build HVAC systems

Create air loops, coils, fans, terminals, schedules, and VAV reheat systems in the right order.

◌

Simulate with confidence

Prepare and run simulations with validation steps that surface issues before they become results.

↗

Turn results into action

Query SQL-backed results, interpret performance, and identify the next modeling decision.

DOE multi-lab collaborationOpenStudio AI brings together National Laboratory of the Rockies, Lawrence Berkeley National Laboratory, Pacific Northwest National Laboratory, and Oak Ridge National Laboratory to deliver one agentic energy-modeling experience.

OpenStudio-MCP integration

Specialized OpenStudio capabilities, ready for your agent.

OpenStudio-MCP gives AI agents a broad set of building-modeling and simulation capabilities through a ready-to-run local Docker image.

◉

OpenStudio AI, powered by OpenStudio-MCP

OpenStudio-MCP brings 150+ tools and 13 workflow skills for model creation, geometry and HVAC work, measures, simulations, and results analysis—all through plain-language MCP workflows.

▣

Local OpenStudio

A ready-to-run Docker image provides a containerized execution environment for the modeling and simulation work your project requires.

⌁

A connected engineering workflow

Model changes, simulation artifacts, decisions, and results remain part of one guided workflow from the first prompt to final review.

＋

More laboratory MCPs are coming

Additional national-laboratory MCPs are being developed for OpenStudio AI, expanding capabilities while keeping one familiar modeler experience.

Use OpenStudio-MCP when a local provider workspace is available. OpenStudio AI also supports a complete local workflow without it. Read the OpenStudio-MCP installation guide or view the Docker Hub image.

OpenStudio AI · guided workflow
You › Add a VAV reheat system to this model.

OpenStudio AI › I’ll inspect the model, create the air loop,
then add terminals, coils, controls, and validation.

  ✓ Model topology inspected
  ✓ Air loop and supply fan created
  ✓ Heating and cooling coils connected
  ✓ VAV terminals configured
  ✓ System validation passed

You › Run the simulation and summarize the results.

  ✓ Simulation complete · Results ready to query
Featured workflow

Prompt to validated VAV reheat system.

Instead of asking an agent to improvise a complex HVAC model, OpenStudio AI coordinates the dependent steps and checks the work before simulation.

1 Inspect the existing model
2 Build the connected HVAC components
3 Validate the system
4 Simulate and query results
The change

Less orchestration overhead. More modeling momentum.

Keep expert judgment in the loop while removing the repetitive coordination work.

Before

  • Hunt through SDK methods and object dependencies
  • Manually coordinate every HVAC component
  • Discover configuration problems after simulation
  • Interpret results in a separate, disconnected step
→

With OpenStudio AI

  • Use reviewed, reusable OpenStudio knowledge
  • Follow explicit workflows and model state
  • Validate before committing to a simulation run
  • Continue from results to the next decision
Quick start

Two steps. One guided modeling experience.

First install OpenStudio AI in the AI client your team uses. Then, if your project needs it, add OpenStudio-MCP as an optional local provider.

Step 1 · Install OpenStudio AI

Codex

Install in Codex

1

Add the marketplace

codex plugin marketplace add <path-to-this-marketplace-folder>
2

Install the plugin

In the Codex plugin UI, install openstudio-ai from openstudio-ai-local.

3

Set up the runtime

In a new task, ask Codex to run setup-openstudio-ai.

Claude Code

Install in Claude Code

1

Add the marketplace

/plugin marketplace add <path-to-this-marketplace-folder>
2

Install and reload

/plugin install openstudio-ai@openstudio-ai-local
/reload-plugins
3

Set up the runtime

Run /openstudio-ai:setup-openstudio-ai in Claude Code.

Step 2 · Optional

Add OpenStudio-MCP

OpenStudio AI works with the plugin alone. Add OpenStudio-MCP when you want its local OpenStudio capabilities; Docker Desktop must be installed and running first.

macOS with Homebrew

brew install --cask docker
open -a Docker
docker version

Windows PowerShell

winget install -e --id Docker.DockerDesktop
docker version
1

Pull the OpenStudio-MCP image

Download the ready-to-run image from Docker Hub.

docker pull nrel/openstudio-mcp:dev
2

Connect it to your AI host

Configure the local provider workspace and restart your client. The setup guide includes the scoped Docker mounts and host configuration.

OpenStudio-MCP setup guide ↗
Start with: “Inspect this OpenStudio model, add a VAV reheat system, run the simulation, and summarize the results.”
Workflow library

Specialized skills for the decisions that matter.

OpenStudio AI combines a shared MCP interface with focused, composable skills—then connects compatible provider backends under the same workflow.

▣

Model editing & SDK guidance

Inspect models and apply durable OpenStudio SDK patterns for geometry, schedules, systems, and more.

openstudio-sdk-model-editor
≋

HVAC system creation

Build fans, coils, air loops, outdoor-air systems, VAV terminals, sizing, and validation workflows.

HVAC workflow skills
◴

Simulation & results

Prepare models, run simulations, and retrieve SQL-backed results without leaving the workflow.

simulate · query-results
✓

Workflow state & repair

Track decisions, validate system state, and repair common issues before the next workflow phase.

state · repair · doctor
⊹

Measures & learning

Propose candidate measures and capture repeatable lessons from modeling sessions.

propose-measure
◉

OpenStudio-MCP modeling

Connect a compatible OpenStudio-MCP provider to the same guided modeling workflow, from preflight through engineering handoff.

OpenStudio-MCP provider workflow
OpenStudio AI for Codex and Claude Code

One place to begin your agentic energy-modeling workflow.

Install the plugin, set up the local runtime you need, and keep a consistent engineering experience as new provider MCPs join the ecosystem.

Choose your client →View on GitHub ↗