Turn Slack threads into shared agent control rooms

What it does

OpenTag is a Node.js gateway that plugs AI agents into Slack. Mention @OpenTag in a thread, and it routes the conversation to local or remote runtimes like Claude Code, Codex, OpenCode, Docker containers, or custom CLIs. The agent replies in the same thread, so the work stays tied to the original team discussion.

The interesting bit

The project treats Slack channels as first-class execution environments rather than notification pipes. Each channel can carry its own project context, memory, runtime defaults, and approval rules—so different teams or projects operate under different safety boundaries without spinning up separate infrastructure.

Key highlights

• Supports multiple agent backends through a pluggable runtime adapter system including mock, Codex, Claude Code, OpenCode, Docker, HTTP, and generic CLI agents
• Versioned “Team Knowledge” with workspace scoping, source references, and revision history that persists across threads
• Channel Profiles and Access Bundles enforce per-channel permissions, allowed runtimes, filesystem roots, and run limits
• Requires human approval for write-capable or risky actions before execution
• Keeps local audit records of sessions, messages, approvals, runs, and artifacts for post-hoc review

Caveats

• Explicitly labeled an MVP; the authors note it is not yet a hosted production SaaS and currently targets local trials and internal experiments
• Local storage handles sessions and audit records, with multi-instance storage and a web admin UI still on the roadmap
• Slack OAuth installation and hardened sandboxing are planned but not yet implemented

Verdict

Worth a look if your team already lives in Slack and wants to share agent work without surrendering accountability. Skip it if you need a managed, production-ready SaaS today.