Skip to content
NagentNagent
Log inSign upHire your AI team
Atlas/Claude Tag
Vendor dossier / Horizontal AI

Claude Tag

Shared Claude teammate that works in Slack channels, with admin-scoped tools, memory and spend limits.

REVIEWED 4 OCTOBER 2026Marketing / Sales / CXNO HANDS-ON BENCHMARK

What it is

Official-source notesHorizontal AIemerging

Anthropic announced Claude Tag on 23 June 2026 as a beta for Claude Enterprise and Team customers, starting on Slack, where one Claude identity per channel can be tagged by anyone in that channel. Claude builds channel memory, can learn from other channels and data sources when granted permission, and can schedule tasks and pursue a project over hours or days. Tasks run in an ephemeral sandbox, and admins extend what Claude can reach through connections, plugins and skills configured per channel, workspace or organisation. With the Respond automatically setting on, Claude may reply to untagged channel messages it judges warrant a reply. Claude Tag replaced the earlier Claude in Slack app, and the product page lists Microsoft Teams as coming soon, with a waitlist. [V1] [V2]

Documented means the statement is present in the cited material. Capability effectiveness, customer outcomes and security assurances have not been independently audited here.

Promoted use cases

  • Triage requests in an intake channel and route what the docs do not cover
  • Answer data questions with warehouse queries and charts in the thread
  • Recap campaign channels, draft announcements and post a weekly campaign metrics digest

Horizontal platforms compete on how people connect company knowledge, delegate work and keep control across tools. The central question is whether a reusable workspace can become an operating process, with a named owner and a reliable path from request to approved action. A visually appealing agent builder is only one part of that decision.

Pricing and buying model

Published commercial evidence

Pricing not published

Available in beta on Claude Team and Enterprise plans. Channel, thread and group DM work bills by usage to an organisation usage balance in its billing currency, capped by a monthly spend limit an Owner sets; adding Claude to Slack adds no per-seat charge, and the documentation states there is no published per-task cost guidance. One-to-one DMs from members with a connected Claude account bill to that member's seat. Team plans must fund the usage balance before Claude responds in channels. A one-time launch credit of USD 25,000 per Enterprise and USD 2,500 per Team organisation expired on 1 October 2026.

Check current commercial source ↗

Budget for implementation, model and tool usage, data, human review and ongoing support. Credit units and outcome definitions differ between vendors. A missing numeric rate is marked unverified rather than replaced with an old third-party estimate.

Collaboration and governance

The documentation describes Claude acting under its own service accounts, with admin-provisioned Access bundles scoped to channels or workspaces, organisation and per-channel spend limits, and an admin Activity page listing scheduled work, memory files (which Owners can edit or delete) and an hourly export of outbound network events. Use of a member's personal connectors in a channel requires that member's approval, with an option to review results before they post. During the beta there is no automatic retention period for Claude Tag data, and the healthcare guide states that Claude Tag is not covered by Anthropic's Business Associate Agreement. The launch post says admins can view a log of everything Claude has done and who requested each task, while the audit documentation states there is no per-action log of every task and who asked; confirm which applies to your plan and how actions are traced through each connected tool's audit log.

The relevant unit of control is permission to update the CRM and release approved follow-up. Inspect who can propose, approve, execute, interrupt and audit that action. Shared seats, shared content and a shared live agent session are different capabilities; require a demonstration of the one your process needs.

Evaluate memory correction, permission revocation and release control for updated instructions. An improvement loop should preserve the original evidence, proposed change, test results and named approver. A safety or security badge alone cannot establish those workflow properties.

Review and customer evidence

No independent review sample was verified in this research pass. Vendor-hosted testimonials are selection-biased customer evidence, not an aggregate rating.

Before relying on a testimonial, confirm the exact product, version, package, workload and baseline. Ask a relevant customer about setup effort, failed cases, ongoing manual work and support after launch. This dossier does not convert customer logos or vendor-hosted awards into independent proof.

Visual source

A screenshot was not captured for this entry. Open the official visual source ↗. The absence of an image does not affect the evidence status of the sourced notes.

Fit, limitations and proof requests

Editorial assessment

Claude Tag places one shared, general-purpose Claude teammate in Slack channels, and its documentation presents marketing, deal and support work as recipes that run on the connections an organisation configures. Beside an end-to-end growth team, it sits as a horizontal collaboration layer whose growth reach depends on those connections, skills and channel instructions.

Priority question: For a task that changes a CRM record or sends a customer-facing message, which approval step can an admin require before Claude acts, and how is that action traced back to the person who asked?

The evaluation owner should be a growth operations lead. Use accepted work completed across tools as the business target. Judge the solution in its own stack role: a cloud runtime, a specialist production tool and a managed AI team can be complementary purchases.

Use-case evaluation plan

Account research to approved follow-up

Give the team an account record, a current product brief and a small set of approved customer references. Ask it to identify a relevant problem, distinguish observed facts from inferred needs, and prepare a follow-up for review. Then change one source and require a second person to correct the draft without restarting the research.

Measure: source accuracy, reviewer minutes, approved follow-ups and CRM consistency.

Campaign brief to coordinated delivery

Start with one campaign objective, audience, budget constraint and brand guide. Require research, a brief, assets and a release checklist with distinct human owners. Introduce an unresolved claim midway through the process. The system should retain useful work while preventing that claim from silently propagating into every asset.

Measure: accepted assets, revision burden, approval latency and missing handoffs.

Customer insight to shared learning

Provide anonymized support themes and sales objections. Ask for a synthesized insight, a proposed campaign adjustment and a knowledge update. A marketing owner and a CX owner should review the same underlying evidence. The exercise tests cross-functional context and accountability, not merely summarization quality.

Measure: evidence coverage, disagreement resolution and reuse of approved learning.

Detailed comparison

Read the detailed Nagent vs Claude Tag guide for operating models, shared context, governance, cost, evidence gaps and a staged pilot.

Nagent vs Claude Tag

Sources and evidence register

Nagent · Multiplayer AI for growth teamsResearch: 21–29 September 2026 · Public-source analysis, no hands-on benchmark · Sources & method