Open Tag: Open-Source Alternative to Claude Tag
Model-agnostic AI teammates in Slack & Teams — built on CopilotKit and the AG-UI protocol
← Part of the Agentic Commerce knowledge base
TL;DR: Just days after Anthropic launched Claude Tag (tagging @Claude in Slack as a persistent team member), CopilotKit shipped Open Tag — a more flexible, open-source version that works with any model, any agent harness, and even fully custom agents.
Source
Atai Barkai (CEO of CopilotKit) posted the launch. The project responds directly to the new wave of embedded AI teammates.
What is Claude Tag?
Anthropic’s recent launch lets teams add Claude directly into Slack channels (with controlled access). You @Claude it like any teammate to delegate tasks, and it works with tools, codebases, and context. It replaces the older Claude in Slack app for Team/Enterprise users.
It makes AI feel like a real team member embedded where work already happens.
Introducing Open Tag (by CopilotKit)
Open Tag brings the same “tag an AI teammate” experience but without vendor lock-in.
Key differentiators:
- Any model — Claude, GPT, Gemini, local/open-source, or whatever you prefer
- Any agent framework or custom agents — not locked to one provider's harness
- Full thread context
- Streaming replies
- Generative UI — charts, cards, forms, interactive elements directly in Slack/Teams
- Human-in-the-Loop (HITL) approvals
- Supports Slack and Microsoft Teams today
- Discord, Google Chat, WhatsApp, and more coming soon
It’s built on CopilotKit’s AG-UI protocol (an open protocol for agent-frontend interactions) and is heading toward full open-source (MIT license for core SDKs; managed/self-hosted enterprise options available).
Demo Highlights
From the announcement video and thread:
- Triage GitHub issues directly in a Slack channel
- Generate breakdowns, charts, and tables
- Create Linear tickets with built-in approvals
- Agent renders rich UI elements and handles collaborative workflows in chat
The experience feels native while giving you full control over the underlying model and agent logic.
Why This Matters (Especially for Sovereign & Agentic Commerce Builders)
This aligns perfectly with the model-agnostic, sovereign approach we advocate for.
In our analysis of Claude Tag in Slack & Notion Agents, one of the biggest risks was context lock-in — your team's operating memory and workflows getting trapped inside a single vendor’s integration. Switching models later becomes painful because the accumulated context lives in their system.
Open Tag directly addresses this by being:
- Model & agent agnostic
- Generative UI + approvals in chat without forcing a specific backend
- Self-hostable / enterprise-grade options
- Built on an open protocol (AG-UI)
- Extensible to more platforms
You can run your own agents (or fleets via tools like PowerLobster, TRAE, or custom orchestrators) and bring them into team chat surfaces with rich interactions — without giving up control of your context layer or being forced into one provider.
Our Take
Tools like this are exactly what the ecosystem needs right now: fast responses to big-vendor moves that prioritize openness and portability.
Open protocols like AG-UI make it easier to build truly portable agentic systems. We can keep owning the important parts (memory, orchestration, business logic) while still getting delightful collaboration experiences inside Slack, Teams, and beyond.
It’s a strong counter to the “rent your company’s brain” risk we discussed with Claude Tag.
How to Get Started
- Request early access via the form linked in the announcement tweet (public launch coming very soon).
- Built on CopilotKit — check out copilotkit.ai for SDKs, docs, and AG-UI protocol details.
- GitHub repo and full open-source components expected soon (core SDKs are MIT).
Related Pages
- Claude Tag in Slack & Notion Agents — The original announcement, productivity wins, and our concerns around context and vendor lock-in. Open Tag is a promising alternative.
- Agentic Commerce — Broader work on sovereign agents, open standards, and infrastructure you control.
- Company Brain — Building unified intelligence and agent fleets without lock-in.
- Mike's CEO Operating System — Structuring work and agent collaboration in portable systems.
- Harness: Practical AI Coding Tips — Staying effective while avoiding hype and dependency.
- AI Operating System: Building a Company OS with Claude and Agents
- Dreaming: Self-Improving AI Agents — Background memory and continuous improvement.