Flux in Your AI Tools: Use the Flux Agent From Claude Code, Cursor, and Any MCP Client

It is Tuesday afternoon. Maya writes the firmware for a soil-moisture sensor. She works in Cursor, and her AI agent writes most of the driver code. The board is in Flux, and her hardware lead designed it last week.

Maya asks her agent: "Write the pin definitions for the sensor board." Before, this was the slow part. She opened the schematic, read each net, typed the pins into a header file, and hoped that nothing had changed since the last export.

Now Cursor is connected to Flux. Her agent finds the project, reads the nets and the components, and writes the pin definitions from the real design. Then it sees a problem: the humidity sensor and the light sensor have the same I2C address. Her agent asks the Flux agent to change the address strap on the humidity sensor and to update the schematic. Flux makes the change in the cloud. In the same time, Cursor continues with the driver. When Flux is done, Maya's agent reads the nets again and updates the address in the code.

The firmware and the board now agree. Maya did not export a file, send a screenshot, or wait for a reply. Her hardware lead opens the project and sees the change, and the message that asked for it.

Flux, in the AI tools you already use

Many engineers already work with an AI agent all day. It writes code in Cursor, plans the week in ChatGPT, or runs long jobs in Claude Code. Until now, that agent could not see the hardware. The schematic, the layout, and the BOM were in Flux, and the agent had no way in.

Flux now has its own MCP server. Connect it to the AI tool that you already use. Your agent can then find your Flux projects, read the design, and hand hardware design jobs to the Flux agent.

What your agent can do

• Find your projects. List and search the projects in your account and in your organizations.
• Read the design. Get the components, the nets, the BOM, the layout, and the project files. Ask questions about a project and get answers from its data.
• Hand design jobs to Flux. Start a new board, ask for the next change on an existing project, or ask for a design review. The Flux agent does the hardware work, and your agent waits for the result.
• Follow the work. Read the design threads of a project and the latest messages from the Flux agent.
• Keep your team's knowledge current. Review the knowledge that Flux suggests from your work, and accept the items that are correct.

The design work runs in Flux, in the cloud. Your agent gets the result. You can open the same project in Flux at any time and see every change and every message.

One connection, every role on the team

Maya is one example. Here is how other people on a hardware team use Flux from their own AI tools.

The hardware lead: a design review before every build

Daniel leads the hardware team. Before each prototype order, he asks Claude Code: "Ask Flux for a design review of every board in the sensor family, and give me one list of the open problems, sorted by board." His agent starts a review in each project. Flux checks the designs, and Daniel gets one list instead of five tabs. He fixes the big items first, and asks Flux for the small ones in the same session.

The project manager: the weekly status writes itself

Priya writes the weekly update for the product team in Notion. She asks Notion AI: "Add the status of the sensor board: the part count from the BOM, the design threads from this week, and what Flux changed." Notion AI reads the project and writes the section. Priya reads it, adds one sentence, and sends it.

Purchasing: the quote request comes from the real BOM

Tom buys the parts. In ChatGPT, he asks: "Read the BOM of the sensor board, and make a quote request table with the manufacturer part numbers and the quantities for 500 boards." The table comes from the BOM in Flux, not from an old spreadsheet. When a supplier says a regulator has a long lead time, Tom asks his agent to have Flux suggest a replacement that fits the board.

The mechanical engineer: the board data for the enclosure

Lena designs the enclosure. She asks her agent for the board size, the mounting holes, and the positions of the connectors and the LEDs. Her agent reads the layout from Flux. When Lena needs the USB connector 2 mm closer to the edge, her agent asks Flux to move it, and she sees the change in the project.

A new product: from a chat to a first board

Sam has an idea for a small battery-powered tracker. He describes it in Claude.ai: the battery, the GPS module, the charging, and the size. Then he asks: "Create a Flux project for this, and have Flux design the first board." His agent creates the project and hands Flux the design job. Flux plans the board, picks the parts, and draws the schematic. Sam opens the project in Flux and continues from there.

The whole team: knowledge that keeps up

At the end of the week, Daniel asks Devin to review the knowledge that Flux suggested from the team's work. Devin lists the suggestions, and Daniel accepts the rules that are correct. The next board starts with the same standards.

Prompts to try

• "List my Flux projects and summarize the BOM of the newest one."
• "Read the nets of my motor controller board, and write the pin map for the firmware."
• "Ask Flux to review the power section of my board, and tell me what it found."
• "Ask Flux to add a reverse-polarity protection circuit to the battery input."
• "Which parts in the BOM of my sensor board have only one manufacturer?"
• "Create a Flux project for a USB-C powered LED controller, and have Flux design the board."

Set it up in a minute

In Flux, open Settings, then AI Integrations, then the Flux MCP tab. Pick your client and follow the steps. For Claude Code, it is one command:

claude mcp add -t http flux https://flux.seobdtools.com/__app__/mcp

Then run /mcp in Claude Code and sign in with your Flux account. Most clients use the same sign-in, so you do not copy a key. If you prefer an API key, or you need access that is restricted to one organization, the MCP server documentation tells you how.

The setup page has steps for Claude Code, Claude.ai, Cursor, VS Code, Codex, ChatGPT, and Notion AI. Devin and other clients that support MCP also work.

Your access stays the same

Your agent signs in as you. It sees only the projects that you can see, and it can change only the projects that you can change. Every job that it hands to Flux shows in the project, like a job that you started yourself.

Try it

Connect Flux to your AI tool, and start with a question about a project that you know well. Then hand Flux a real design job, and keep working while it runs.