OpenROAD MCP: chip physical design from an agent
OpenROAD MCP Server
MCP server from the OpenROAD project: an agent opens OpenROAD sessions, runs commands, checks history, metrics and ORFS reports
- 35 ★ on the repository
- Last commit
- BSD-3-Clause
Extended access
Why this level
- An agent runs OpenROAD commands in real sessions and may change project files
- Sessions left open keep consuming machine resources
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/openroad-mcp-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- An agent runs OpenROAD commands in real sessions and may change project files
- Sessions left open keep consuming machine resources
This is third-party code. Review the repository files before installing.
What it does
The server connects an agent to OpenROAD, an open toolset for physical design of digital chips from RTL to GDSII, and to the OpenROAD-flow-scripts flow. It starts interactive OpenROAD sessions with full PTY support, and the agent creates, lists, inspects and terminates several sessions at once. In a session the agent runs commands and queries and reads the full command history and performance metrics. Separate tools show ORFS flow reports: they list report images and read them right in the chat. There are ten tools in all. The server runs through npx -y openroad-mcp or a Docker image and is in the MCP registry. OpenROAD itself and optionally ORFS must be installed beforehand, and the server finds them on PATH and in usual install locations.
Who it is for. Chip design engineers and students who work in OpenROAD and ORFS and want to run sessions and go through reports together with an agent.
Good fit when
- You run a project in OpenROAD and want the agent to execute commands in a live session
- After an ORFS flow run you want to open report images and go through the result
- Several experiments run in parallel and each needs its own session
- You want to recall which commands ran in a session and look at the metrics
- You walk through a learning RTL to GDSII route with an agent step by step
Not a fit when
- OpenROAD is not installed on the machine: the server only connects an agent to it
- You need a digital logic project without physical design: the server is for the OpenROAD flow
- The project is confidential and the agent must not run session commands: the tool runs them for real
- Node.js is older than 22: the npx package will not start on it
Example request
Open an OpenROAD session for the current design, run the placement step, show the metrics and open the report images of the last ORFS run.MCP
- Transport
- stdio
- Authentication
- not required
Environment variables
ORFS_FLOW_PATH- Path to the flow directory of OpenROAD-flow-scripts
Limitations
You need Node.js 22 or newer and OpenROAD installed on PATH. ORFS is optional but recommended for the full RTL to GDSII flow and report viewing. The ORFS path defaults to ~/OpenROAD-flow-scripts/flow and is changed with the ORFS_FLOW_PATH variable. The npx package does not read a .env file.
A client launched from a GUI may not find openroad on PATH, and then the path is passed through an environment variable. The tools run OpenROAD commands in real sessions with access to project files, and detailed parameters are in docs/API.md.
How to disable. Remove the server with claude mcp remove openroad-mcp. Running sessions end with the terminate_interactive_session tool or together with the client process.
FAQ
How do I connect OpenROAD MCP to Claude Code?
With claude mcp add --transport stdio openroad-mcp -- npx -y openroad-mcp. OpenROAD must be installed and on PATH, or the server cannot start a session. There is a Docker image, ghcr.io/the-openroad-project/openroad-mcp.
Is ORFS required?
Not required, but recommended. With ORFS the agent runs the full RTL to GDSII flow and reads report images. Without it the server still runs OpenROAD sessions, executes commands and shows history and metrics.
Which tools does the server have?
Ten: execute and query in an interactive session, create, list, inspect and terminate sessions, command history, session metrics, a list of report images and reading an image. Parameters are in the repository's docs/API.md.
What if the client cannot find openroad?
Pass PATH with the openroad directory and the ORFS_FLOW_PATH path as environment variables when adding the server. The server fills PATH from the login shell and usual places itself, but GUI clients may start with a trimmed environment.
Does the server handle several sessions?
Yes, the agent creates several interactive sessions, lists them, inspects their state and terminates them one by one. Each has its own command history and performance metrics.
Discussion
No questions or reviews yet.
Related
MCP serversOpenMCP turns any OpenAPI specification into an MCP server and lets you choose which tools the agent sees
The openmcp CLI and a VS Code extension for debugging MCP servers: tools, prompts and resources, chat with a model, test cases
The MCP server turns OpenAPI operations into agent tools and sends real requests to the configured REST API.
An MCP server exposes Rails project structure, routes, models and schema to agents alongside framework documentation.
An MCP server that lets an agent read files, search code and run commands in a workspace
A plugin with skills for building TypeScript MCP servers and MCP apps: the agent scaffolds, edits, debugs and verifies them
README in short
The README explains the server's purpose, shows a demo and lists requirements: Node.js, OpenROAD and optionally ORFS. Then come path setup, collapsible instructions for many clients, a list of ten tools and a troubleshooting section.