Agent Console
wangqioo/agent-console
Controlador local de hardware para agentes Claude e Codex via API HTTP
Ver no GitHubDaemon local que funciona como controlador de hardware para gerenciar sessões Claude Code e Codex CLI. Monitora status de provedores, inicia/controla sessões de agentes, e expõe uma API HTTP simples para integração com dispositivos físicos como botões e displays.
Instalação
npx convez-ai wangqioo/agent-console1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Integrar botões físicos para controlar sessões de IA
- Monitorar status de múltiplos agentes Claude/Codex simultaneamente
- Automatizar fluxo de trabalho com controladores de hardware
- Gerenciar sessões CLI através de interface REST
Veredito

Projeto experimental bem estruturado para quem busca controlar agentes IA via hardware. API limpa e exemplos práticos, mas ainda em estágio de protótipo — requer conhecimento técnico de integração de hardware e CLI.
README
A tradução para o português deste README ainda está na fila.
Agent Console
Agent Console is a local daemon prototype for a desktop hardware controller that can monitor and control Claude Code and Codex CLI sessions.
The first milestone is intentionally small:
- query Claude and OpenAI/Codex service status from public status APIs
- start local CLI sessions
- read recent session output
- send input to a session
- provide a simple
continuecommand for a hardware button
Reference Projects
These two repositories are the primary references for this prototype:
graykode/abtop- https://github.com/graykode/abtop- Reference for discovering and monitoring AI coding agent sessions, including Claude Code and Codex CLI.
- Useful ideas: session list, token/context/rate-limit visibility, process and port monitoring.
alexei-led/ccgram- https://github.com/alexei-led/ccgram- Reference for controlling Claude Code and Codex CLI through terminal sessions.
- Useful ideas: bridge to terminal sessions, monitor output, send replies, manage multiple sessions.
Prototype API
Run the daemon:
python .\daemon\server.py
Default URL:
http://127.0.0.1:8765
Endpoints:
GET /api/providers/status
GET /api/agents
GET /api/agents/{id}
POST /api/agents/{id}/start
POST /api/agents/{id}/send
POST /api/agents/{id}/continue
POST /api/agents/{id}/stop
GET /api/sessions
POST /api/sessions/start
GET /api/sessions/{id}
POST /api/sessions/{id}/send
POST /api/sessions/{id}/continue
POST /api/sessions/{id}/stop
Example:
Invoke-RestMethod http://127.0.0.1:8765/api/providers/status
List configured agents:
Invoke-RestMethod http://127.0.0.1:8765/api/agents
Start Codex from config.json:
Invoke-RestMethod http://127.0.0.1:8765/api/agents/codex/start -Method Post
Continue Claude:
Invoke-RestMethod http://127.0.0.1:8765/api/agents/claude/continue -Method Post
Start a session:
Invoke-RestMethod http://127.0.0.1:8765/api/sessions/start `
-Method Post `
-ContentType application/json `
-Body '{"name":"shell","command":["powershell","-NoProfile"]}'
Configured Agents
Edit config.json to point agents at the projects and commands you actually
use:
{
"agents": [
{
"id": "claude",
"name": "Claude Code",
"command": ["claude"],
"cwd": "C:\\Users\\100448405\\your-project"
},
{
"id": "codex",
"name": "Codex CLI",
"command": ["codex"],
"cwd": "C:\\Users\\100448405\\your-project"
}
]
}
Send input:
Invoke-RestMethod http://127.0.0.1:8765/api/sessions/1/send `
-Method Post `
-ContentType application/json `
-Body '{"text":"Write-Host hello\n"}'
Hardware Direction
The daemon should expose a stable local API. The hardware side can then be any small device that speaks USB serial, WebSocket, or HTTP through a companion bridge.
Initial controls:
- agent switch button
- continue/confirm button
- cancel/stop button
- small screen for provider status and active session output