LabWired Agent
Write firmware. Prove it on a digital twin — and on a real board when you plug one in.
macOS / Linux
curl -fsSL https://labwired.com/install | bash
labwired agent
Windows (PowerShell)
irm https://labwired.com/install.ps1 | iex
labwired agent
Paste the command for your OS into a terminal. Then run
labwired agent.
·
source
Write. Check. Fix. Green.
Twin first (deterministic). Real board when you have silicon. Honest claims either way.
-
01
Write
Describe the board and the task.
-
02
Check
Runs on the digital twin (or your board).
-
03
Fix
Wrong output. Patch and re-run.
-
04
Green
model_verifiedon twin;hardware_observedon desk.
Your agent, or ours.
Same twin tools either way. Desk hardware is local.
Free, Hobby, and Pro AI credits are shared across browser Architect and hosted CLI agent. BYO Codex, Claude, or local model traffic is distinct from LabWired-hosted inference.
LabWired Agent
CLI + skills. Local model optional.
curl -fsSL https://labwired.com/install | bash labwired agent
Windows: irm https://labwired.com/install.ps1 | iex
Claude, Codex, Cursor, VS Code, OpenCode
MCP into the same tools.
curl -fsSL https://labwired.com/mcp.sh | sh
Windows: irm https://labwired.com/mcp.ps1 | iex · manual: claude mcp add labwired --transport http https://api.labwired.com/mcp
Included.
Questions.
Do I need a board?
No for the twin path — prove on a digital twin first. When a board is plugged in, use desk-hw (probe flash + serial marker → hardware_observed). Twin green and hardware green are separate claims.
Does the model grade itself?
No. Only labwired_verify mints model_verified on the twin. Compile success and “looks right” are not enough.
Claude or Codex?
Yes. MCP uses the same tools as the CLI agent.