Anthropic has announced Claude Tag, a new collaborative AI experience that lets teams assign work to @Claude directly inside Slack. Claude Tag transforms Claude from an individual chatbot into a shared AI teammate that can collaborate across channels, remember context, access approved tools, connected data and codebases, and complete long-running tasks.
Anthropic says Claude Tag represents the next evolution of Claude Code, making the AI more proactive and better suited for team collaboration. According to the company, around 65% of its product team’s code is now created using its internal version of Claude Tag.
Claude Tag brings a shared AI teammate to Slack
Claude Tag enables anyone in an authorized Slack channel to mention @Claude and delegate work without leaving Slack. Claude understands the full context of a thread, breaks tasks into multiple stages, posts a checklist while it works, and shares the final result in the same conversation.
Unlike traditional AI assistants that work in individual chats, each Slack channel gets a shared Claude that acts as a teammate for everyone in the conversation. Everyone can see what Claude is working on, continue conversations started by others, and build on previous interactions without repeating context.
Teams can use Claude Tag to:
- Summarize long Slack discussions
- Turn decision threads into documents
- Reproduce software bugs and draft pull requests
- Query business metrics and reports
- Investigate support tickets
- Prepare meeting briefs
- Monitor channels, triage alerts and surface work that needs human review
- Organize project updates
Users can also send Claude direct messages, which use their personal Claude account and connected tools instead of the shared channel configuration. Tasks performed in shared Slack channels are billed to the organization, while direct messages use the individual’s Claude account and personal connectors. Organizations can also disable direct messages if required.
Memory, proactive assistance and long-running tasks
Claude Tag continuously builds context from the Slack channels it has permission to access, allowing it to understand previous discussions, project history, team decisions, ownership information and organizational workflows over time.
If administrators grant permission, Claude can learn from:
- Slack channel history
- Previous discussions
- Team decisions
- Connected data sources
Anthropic says Claude only accesses channels and connected data sources it has permission to use, and it does not report information from private channels unless authorized.
Anthropic has also introduced an optional ambient mode, allowing Claude to work proactively instead of waiting for prompts.
With ambient mode enabled, Claude can:
- Monitor conversations
- Surface important updates
- Highlight unresolved discussions
- Follow up on inactive threads
- Notify teams about relevant developments
- Tag users when decisions require attention
Claude Tag is designed for asynchronous work, allowing teams to delegate complex projects while focusing on other priorities. Teams can also give Claude standing instructions to monitor channels, generate recurring updates, watch for urgent activity, or perform scheduled tasks.
Claude can also:
- Schedule future tasks
- Continue working independently for hours or days
- Request clarification when needed
- Return with completed results
- Handle multiple delegated tasks in parallel
Enterprise tools and integrations
Claude Tag connects to enterprise applications through Agent Identity, allowing it to securely access approved tools, query data, perform actions, and post results directly back into Slack threads.
Supported integrations include:
- GitHub
- HubSpot
- Atlassian
- Datadog
- Salesforce
- Figma
- PagerDuty
- Zendesk
- Linear
- Asana
- Snowflake
- BigQuery
- Gong
Anthropic says organizations can also connect additional API-based tools.
Admin controls and setup
Claude Tag has been designed with teams and organizations in mind. By default, Claude Tag has no access to external systems until administrators configure approved credentials, repositories, tools and other resources.
Organization administrators decide which Slack channels Claude can access, which external tools and data sources are available, and where each Claude identity can operate. Claude’s access to tools and data depends on the permissions configured for each Slack channel rather than the individual requesting the task.
Organizations can create separate Claude identities for different teams, ensuring each identity maintains its own memories, permissions and connected resources. For example, a Claude configured for engineering cannot access sales tools, data or memories unless administrators explicitly allow it.
Claude performs actions using its own identity instead of individual user credentials, while every credential use is logged so administrators can review what Claude did and who requested each task.
Administrators can configure:
- Slack channel access
- Connected tools, repositories and codebases
- Data source permissions
- Separate Claude identities for different teams
- Monthly spending limits for organizations and individual channels
- Activity logs and audit history
Setting up Claude Tag involves:
- Pairing Claude Tag with the Slack workspace
- Granting Claude access to approved tools
- Setting the organization’s monthly spending limit
- Testing Claude Tag in a private Slack channel
Once configured, everyone in authorized channels can begin using @Claude without additional setup.
Availability
Claude Tag is available today in public beta for Claude Enterprise and Claude Team customers using Slack. The new experience replaces the existing Claude in Slack app, with administrators able to migrate within a 30-day window.
Anthropic is also providing eligible Enterprise and Team organizations with introductory launch credits to help teams evaluate the feature. The company says it plans to expand Claude Tag beyond Slack, bringing the shared AI teammate experience to more workplace platforms in the future.