linxidnju/OpenTag

OpenTag sits between your Slack workspace and whatever AI agents your team runs — Claude Code, Codex, OpenCode, Docker-based tools, HTTP agents, custom CLIs — and handles the routing, policy enforcement, approval gates, memory, and audit logging that none of those agents bring out of the box. The real differentiator here is that it is channel-native: agents are invoked from Slack threads, and their outputs land back in context, which is where most engineering teams already coordinate. Compared to the field of generic agent orchestration frameworks, this one has an opinionated scope: team workflows inside Slack, not yet another general-purpose pipeline builder. The approval layer is worth the install alone if your team needs a human-in-the-loop gate before agents push code or trigger deployments. Reservation: the project is early and the documentation will need work before a non-technical PM can self-serve. -> Best for: SaaS team of 2-5 already running AI coding agents who want guardrails and traceability without building the plumbing themselves