Claude Slack

The Problem

Claude Code is incredibly powerful, but it's chained to your terminal. If you step away from your desk — grabbing coffee, sitting on the couch, picking up the kids — you lose access to your AI development environment entirely. For someone who builds with Claude Code daily, that's a lot of dead time where work could be moving forward.

My Role

I designed and built the full system: the wrapper, the Slack integration, and the interaction model.

The Approach

I built a wrapper around Claude Code that bridges it to Slack. When you run claude-slack, it creates a new thread in a dedicated Slack channel and routes all of Claude's output there — messages, permission requests, multi-choice prompts, everything. Your responses in Slack flow back through the wrapper and into Claude Code as if you were typing in the terminal.

The key insight was that Slack already solves most of the hard UX problems: notifications, speech-to-text, mobile-friendly input, and threaded conversations. I just needed to connect the pipes.

What I Built

• Slack thread orchestration — each claude-slack session spawns a dedicated thread, keeping conversations organized
• Emoji-based quick responses — numbered choices from Claude get corresponding emoji reactions; tap to respond without typing
• Speech-to-text input — use Slack's built-in voice messages to talk to Claude naturally from your phone
• Push notifications — get a Slack ping when Claude finishes thinking, so you don't have to watch and wait
• Full local environment — Claude still runs on your machine with all your skills, configs, and project context intact
• Bidirectional message passing — Claude's full output streams to Slack; your Slack replies stream back to Claude

The Result

I used this daily. It turned downtime into productive time -- I can review Claude's work, approve permission requests, and steer development from anywhere with a phone signal. The emoji reaction pattern for quick choices turned out to be surprisingly fluid; most interactions are a single tap. It also meant I got notified the moment Claude needs input, eliminating the "stare at terminal waiting" problem entirely back when that was the status quo. In full candor, this was built before 'claw' was a household term or /rc was available in Claude mobile. I needed functionality that didn't exist, so I built it. Now there are plenty of well-maintained more robust alternatives so I've abandoned this project.

Tech Stack

• Runtime: Node.js wrapper around Claude Code CLI
• Integration: Slack API (Bot, threads, reactions, real-time messaging)
• Input: Text, emoji reactions, Slack voice messages (speech-to-text)
• Environment: Full local Claude Code with skills, configs, and project context