Skip to main content
This guide takes you from installation to a first verified change on your board. You install Embedder, sign in, connect your repository to an Embedder project, record its commands with /init, ask for a small change, and watch the board’s serial output.

Choose how to start

Where a step has no tabs, it’s the same in VS Code and the terminal UI.

Install and sign in

You need VS Code 1.106.0 or newer, or a fork built on it, such as Cursor or Windsurf. See VS Code extension requirements.
1

Install the extension

Open the Extensions view with Cmd+Shift+X on macOS or Ctrl+Shift+X on Windows and Linux. Search for Embedder and install the extension.The extension also installs the matching Embedder CLI. You don’t need to install it separately.
2

Open your firmware folder

Open your repository with File > Open Folder. Embedder works in the folder you open, or in the first folder of a multi-root workspace. Without a folder, chat shows Open a folder to get started.
3

Open chat

Open the Command Palette with Cmd+Shift+P on macOS or Ctrl+Shift+P on Windows and Linux, and run Embedder: Open Embedder Chat. Chat opens in the Secondary Side Bar. In Cursor, it opens as an editor tab in a locked group on the right.
4

Sign in

Select Sign in with Embedder. Chat shows a six-character code, and your browser opens the Connect VS Code page. Enter the code, select Authorize VS Code, and return to VS Code when the page shows You’re set up.If the browser doesn’t open, select Open verification page. If the page asks you to sign in to Embedder first, sign in, and it returns to Connect VS Code.
5

Choose a team

If you belong to more than one team, choose one under Select a team. If you belong to one, Embedder selects it for you.

If setup stops

For anything else, see Get help.

Connect a firmware project

An Embedder project holds your repository’s hardware context: the platform, the peripherals fitted to your board, and any uploaded schematics. Embedder uses it to find the right documentation and board skills. If your team has only one project, or you used this folder with a project before, Embedder selects that project without asking. Run /project to switch to another one.
  1. Under Select a project, choose an existing project, or select New project. To narrow a long list, search or pick a family above it. If your team has no projects yet, Embedder skips this list.
  2. Upload your schematic opens. Create the project in one of two ways:
    • From your schematic. Enter a Project name, add the design files, and choose Parse & Create. See Start a project from a schematic.
    • From the catalog, if you don’t have a schematic yet. Select Pick a platform instead. Under Pick your platform, search for your board or MCU and select it. Choose the specific board when it’s listed, and check the exact part number. Enter a Project name or keep the suggested one, and select Continue. Under What is on your board?, select the sensors, displays, radios, and other parts fitted to your board, then select Create project. Select Skip for now to add them later.
When the project opens, chat shows a Get started list: open or add your schematic, connect to serial, ask for something small, and take the tour. Each item checks off when you do it. Close the list when you no longer need it. It also disappears after you send five prompts.If the platform list shows Couldn’t load platforms, check your connection and select Retry.To rename or delete a project, use the Rename and Delete buttons on its row under Select a project.
If your development board isn’t listed, select its MCU. If neither is listed, as with a custom board, start the project from your schematic. See Schematics for supported formats, and Hardware catalog for working with hardware outside the catalog. To add or review the parts fitted to your board later, run /peripheral. To give Embedder the board design, run /upload-schematics, then check the detected components against your board revision. See Add and manage peripherals for suggestions, custom parts, and removal.

Record the project commands

Run /init to create EMBEDDER.md in the project root. This is the project guide that Embedder follows in every conversation. In VS Code, you can also select Setup Project at the top of a new chat.
Embedder inspects the repository, then asks about five areas, one step at a time: Target hardware, Build & test, Debug & deploy, Bench & integrations, and Quality & rules. Each step shows a step count, such as 1/5, and a progress bar.
  1. In each step, choose the option that matches your project and select Continue. Each step shows what Embedder detected and why.
  2. To correct a wrong target, command, or pin assignment, choose Something else… and type your answer. Typing is optional. Select Back to change an earlier step.
  3. In the last step, select Finish setup.
In the terminal UI, press Enter to continue and Esc to go back. Discovery only reads your project. Embedder builds, flashes, or runs a debugger only for the checks you choose during setup.
Approval prompts don’t appear while /init runs. Choose a flash or debug check only when the board is safe to program, halt, and reset.
Review EMBEDDER.md when setup finishes. See Configure EMBEDDER.md.

Ask a focused question

Start with a question that shows whether Embedder understands the project:
Type @ to point at a file or directory:
Choose a mode with the mode control in the message box, or press Shift+Tab to cycle modes: See Agent modes for Debug and the other modes.

Make and verify a change

Ask for a change and the evidence that proves it:
With the default approval mode, Manually approve, Embedder reads files and runs a reviewable list of read-only commands without asking. It asks before it edits a file, runs any other command, or uses hardware. See Tool approvals. Review each edit after Embedder makes it:
  • In VS Code, the changed blocks are highlighted in the editor. Choose ✓ Accept or × Reject on each block. See Review edits in the editor.
  • In the terminal UI, run /review-edits to accept or reject edits per file or per change.
A successful build is the first check. With the board connected, ask for the runtime result too:
Embedder asks for confirmation before it programs the board, unless you chose Skip all approvals.

Observe the board

Use the Embedder Monitor to watch serial output while you work:
  1. Run /monitor. In VS Code, the Monitor opens in the Panel. In the terminal UI, it opens in your browser.
  2. Select + to open a New tab, then choose Serial Monitor.
  3. Select the port chip at the start of the toolbar. Choose your port under Port, set Mode and Baud, and select Connect.
The tab shows the board’s output as it arrives. See UART for wiring, port settings, and boot output. In the terminal UI, the browser opens on the machine that runs Embedder. Over SSH, run /serial to open the serial sidebar in the terminal instead. Then ask Embedder to read the output or take a bounded measurement:
For an investigation that needs breakpoints or target-state inspection, run /debug to switch to Debug mode. See GDB debugging. If the board has faulted, ask Embedder to attach without programming or resetting it, and read Attach to a running target before you connect.

Continue

Common workflows

Plan, build, test, flash, and verify a change.

Monitor and debug

Choose serial, debugger, or instrument evidence.

VS Code extension

Find chat, edit review, the Monitor, and schematics in the editor.

Use Embedder in the terminal

Work from a terminal or SSH session.
Last modified on November 13, 2026