Your Claude Code session, now answering in Slack from the beach

What it does

Vibe Remote is a local-first relay that puts Claude Code, OpenCode, and Codex behind Slack, Discord, Telegram, WeChat, and Lark. You message your agent from your phone, it works on your actual machine, and you approve or steer it from chat. Each chat thread maps to an isolated agent session, so five threads means five parallel jobs you can resume later from any device.

The interesting bit

It’s deliberately not an agent framework — the middleware adds no reasoning loop of its own, so tokens go to the agent you chose rather than a persona layer. It also ships a small harness of durable primitives (vibe agent run, vibe task, vibe watch, vibe runs) so agents can schedule, monitor, and report back on their own timeline instead of being babysat.

Key highlights

• Local-first: Socket Mode / WebSocket only, no inbound ports or public endpoints
• Web wizard handles setup, dashboard, per-channel agent routing, and health checks
• Interactive prompts surface as buttons/modals in chat — full CLI interactivity, no terminal
• Completion notifications let you delegate and walk away
• vibe remote exposes the local Web UI through a secure avibe.bot tunnel when you’re off-network

Caveats

• The remote-UI tunnel routes through avibe.bot’s infrastructure, which slightly undercuts the pure local-first story
• Windows is effectively a WSL-only affair per the docs
• The README spends a lot of energy comparing against “OpenClaw” — take the token-cost claims as vendor math

Verdict

Worth a look if you already run Claude Code or Codex and want to drive them from your phone or team chat. If you don’t live in chat apps, it’s plumbing for a lifestyle you don’t have.