# O vs Claude Tag

Claude Tag is a per Slack Channel agent, centrally set up by owners. @O gives every employee their own AI coworker in plain English, on any model, across 1,000+ tools.

## TLDR

### O gives every employee their own AI coworker.
- **Setup**  
  @O is a one-click admin install, anyone can use their own AI coworker day one.
- **Model choice**  
  @O runs on any provider, including your own local LLM.
- **Agent philosophy**  
  @O is one AI coworker per user that works and remembers across Slack Channels.
- **Connectors**  
  @O connects 1,000+ tools out of the box, inside Slack.

### Claude Tag is an owner-provisioned, per-channel agent.
- **Setup**  
  Tag is provisioned as an agent per Slack Channel, centrally managed by an owner.
- **Model choice**  
  Tag runs on Claude only.
- **Agent philosophy**  
  Tag is centrally defined by an owner on a per channel logic.
- **Connectors**  
  Tag ships ~14, each owner-wired, extendable via custom MCP.

## Comparison table

| Feature              | @O                                                                                       | Claude Tag                                                                                  |
|---------------------|-------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------|
| Status              | Live.                                                                                    | Public beta. Enterprise and Team plans only.                                             |
| Setup               | Bottom-up. One-click admin install, no API keys. Any employee delegates in plain English, day one. | Top-down. Owner provisions per Slack Channel: pair Slack, wire each tool's auth, set spend limits. Not a 5-minute install. |
| Model               | Multi-model, including your own local LLM for cost control.                               | Claude Opus 4.8 only.                                                                       |
| Agent philosophy    | One AI coworker per user, with persistent context across Slack Channels, tools and time. | Centrally defined by the owner by Slack Channel, context scoped to that Slack Channel. Personal Claude in DMs. |
| Connectors          | 1,000+, added inside Slack in plain English. Custom API connector available.              | ~14 curated at launch, each owner-wired. Extendable via custom MCP an admin sets up.     |
| Agent identity      | Per-user identity. Each AI coworker acts as its user, with attributable actions.           | Acts under its own service account. Actions log as "Claude".                             |
| Proactivity         | Proactive by design. Schedules its own work and suggests automation before you think of it. | Ambient mode exists, opt-in.                                                              |
| Memory              | Portable, inspectable, model-neutral.                                                    | Channel-bound, non-portable.                                                               |
| Pricing             | Metered API-rate usage with a credit system. $100 free credit.                           | Metered API-rate usage. One-time launch credit.                                            |

## When to choose @O vs. Claude Tag

### Choose @O if
- You want every employee using AI on day one, not just engineers.  
- You need model flexibility, including running your own local LLM.  
- You want one AI coworker per user that remembers across tools and time.  
- You want people to connect their own tools in plain English, within guardrails.  
- You need actions attributed to the person who asked for them.

### Choose Claude Tag if
- Your team is technical and happy to own its own setup.  
- You are all-in on Anthropic and don't need model choice, nor vendor flexibility.  
- Your work lives inside a few well-scoped Slack Channels.  
- You already run Claude Enterprise or Team and want to extend it.  
- A shared service-account identity works for your governance model.