Wireface GitHub

Coding

A coding agent you can talk to.

Run Claude Code, OpenCode or Codex behind the face. Say what you want: it works in your repo in the background, asks before it runs or changes anything, and tells you what it found and what it changed.

Get started Watch the demo
A developer at a desk at night; on the monitor, a code editor with the glowing wireframe face floating beside the code and the activity panel on the right

Watch it work

Real agent, real voice, real time

Claude Code finds and fixes two bugs in a checkout module. The face asks before it runs anything, then tells you what was wrong. Recorded in one take, nothing sped up.

Things to say

Ask the way you'd ask a colleague

  • fixRun the tests and fix whatever fails.
  • reviewLook at my last commit and tell me what you'd change.
  • refactorRename UserStore to AccountStore everywhere and keep the tests green.
  • explainWhy does the app crash on startup? (drop the log file on the face)
  • check inHow's it going?

It reads, runs and edits in the folder you point it at, like it would in a terminal, and you hear the outcome: what failed, what it changed, what it needs from you. Code, paths and links are described rather than read out; the details are in the panel.

Ask "how's it going?" in the middle and the voice answers at once from what the agent has done so far. Say its name to cut in and change course.

While it works

You can always see what it's doing

The activity panel

Every request as a timeline: tool calls grouped and expandable, results, subagents, background tasks, and the cost of each request. Claude Code's full transcript is a tab away. Dock it to the screen edge as a drawer.

Permissions, out loud

Claude Code and OpenCode ask before they run a command or change a file. The face asks out loud and shows Allow, Always and Deny; with the microphone on, "yes" or "no" answers. Codex's sandbox mode is chosen up front.

It can use your PC

Claude Code can take screenshots and use the mouse and keyboard, drive a Chrome window of its own through Playwright, or work in your own Chrome: check the page it just built, click through a flow, read an error dialog.

Rooms

Several agents, one conversation

Put faces in the same room and they behave like a team around you. The room owns the microphone, so you are heard once, and only one face speaks at a time.

A developer with a coffee watching three faces on a monitor, a cyan wireframe, a white android and Dracula; the android is speaking
Claudeleads: answers first
Codexadds a point, or passes
OpenCodeadds a point, or passes
  • One round per question. The first agent with something to say leads. Each of the others adds a sentence or two, or passes. Then the lead concludes. Time limits on every turn, so they can't loop.
  • Say a name to choose. "Codex, what do you think?" goes to Codex first. "Everyone, ..." goes to all. Say a name, or "stop", while anyone is talking and they all stop.
  • Faces only hear you. Agents never hear each other's voices. What one says reaches the others as text, labelled with its name.
OpenFace.cmd --face claude   --room team
OpenFace.cmd --face codex    --room team
OpenFace.cmd --face opencode --room team

What it needs

  • Windows 10 or 11, with Python 3.11 or newer and Node.js (the first run installs the rest).
  • Claude Code, OpenCode or Codex, installed and signed in. Your settings, skills and MCP servers come along.
  • Optional: a Gemini or OpenAI API key for the realtime voice. Without one, a voice that runs on your PC reads the replies.

Give your coding agent a face

Clone it, run one command, pick the folder your agent works in, and say hello.

Get startedSee the assistant