What Is Claude Tag? Pricing, Setup, and How It Works
Claude Tag is Anthropic's shared AI teammate for Slack. Anyone in a channel can tag @Claude, hand it a task, and watch it work in the thread, using the tools and data an admin connected for that channel. It launched in beta on June 23, 2026, for Claude Team and Enterprise plans.
A lot has changed since launch week, though: it replaced the old Claude in Slack app in August, it bills by usage on top of your Claude plan, the launch credit has already expired, and Owners now choose which model it runs on.
On October 11, 2026, we went through Anthropic's launch post, the Claude Help Center, the full Claude Tag docs, and the promo terms, so this guide covers how Claude Tag works today, what it costs, who can get it, how to set it up, and the limits worth knowing before you roll it out.
What is Claude Tag?
Claude Tag is a Slack integration that puts one shared Claude into your team's channels as a working teammate. You tag @Claude into a conversation and describe what you need, and it breaks the job into steps, uses the tools your admin connected, and posts the result in the thread.
Unlike the regular Claude app, where the conversation belongs to you, Claude Tag gives the whole channel the same Claude, so a teammate can jump into the thread, add context, or pick up the work where you left it.
It's the same shift we've been watching across the whole AI teammate category, where the assistant lives inside the team's chat and works for everyone in it.
Here are the key facts at a glance:
Anthropic announced Claude Tag on its blog, and The Verge covered the launch the same day, describing Claude joining a company's Slack to write pull requests, find sales numbers, and take on delegated tasks.
{{templates}}
How Claude Tag works
Once an admin sets it up, there's nothing for regular users to install. You type @Claude in a channel Claude has joined, say what you want in the same message, and Claude picks it up.
For a quick question, it just replies. For a bigger job, it posts a checklist in the thread and ticks items off as it works, which is handy to know because Slack doesn't send notifications when a message is edited (so a thread can look frozen while Claude is halfway through the list).
One shared Claude per channel
Every channel has one Claude that everyone works with. According to Anthropic's docs, anyone in the channel can steer a running task by replying in its thread, without tagging Claude again or starting over. If you need it to stop, there's a Stop button on the working indicator.
Three ways to reach it in Slack
The Claude Help Center lists three places you can work with Claude Tag:
- Channel tagging: tag @Claude in a channel and follow along as it works in the thread. This is the shared, team-facing mode.
- Direct messages: start a private conversation with @Claude. In a DM, Claude uses your own Claude account and your own connected tools.
- AI assistant panel: click the Claude icon in Slack's AI assistant header to open a side panel you can use from anywhere in Slack.
Memory that stays in its lane
Claude Tag remembers context per channel and per workspace, so your team doesn't have to re-explain the same background every time someone asks for help. Memory survives between tasks, and Owners can view, edit, and delete it.
Anyone in the channel can ask @Claude what do you remember about this channel? and correct what comes back.
The memory follows Slack's boundaries. Notes saved in a public channel can become workspace notes that every channel reads, while a private channel can read those workspace notes but keeps everything it learns to itself.
It works under its own identity
In channels, Claude works under its own accounts, which an admin provisions tool by tool, so everything it does shows up in each tool's audit log under Claude's name (handy when someone asks who closed that ticket).
Code changes come from the Claude GitHub App as a draft pull request, with a link back to the Slack thread that started them.
The work runs in a temporary cloud sandbox that uses the same engine as Claude Code on the web. Claude can read files, run code, build charts, and open pull requests there, and the sandbox is released a few minutes after the task finishes.
Scheduled work and ambient mode
Claude Tag can run on a schedule and follow up on its own. You can ask it to post a digest every morning, watch a pull request, or chase an approval until it lands, and it can schedule its own follow-up tasks to keep a project moving over hours or days.
There's also an optional ambient mode. When it's on, Anthropic says Claude proactively flags relevant information and nudges threads or tasks that have stalled.
Work Claude starts on its own still bills to your organization, so decide which channels get this behavior before you switch it on.
What you can use Claude Tag for
Claude Tag is built for shared work the whole channel should see, like turning threads into documents, triaging requests, answering data questions, and turning bug reports into draft pull requests. Anthropic built it for its own teams first, and the launch post says 65% of Anthropic's product team's code is now created by its internal version of Claude Tag.
That's Anthropic's own figure about its own engineers, so read it as a sign of how heavily they lean on the tool. Your team's number will depend on what you connect and how you use it.
The more useful examples come from Anthropic's post on how its employees use Claude Tag, published August 28, 2026:
- Turning a messy thread into a document: a product marketer tagged Claude into a Slack thread of more than 15 messages and got a two-page draft back in about two minutes. The full job, across four versions of the doc, took about 45 minutes.
- Rounding up scattered requests: someone on product strategy and operations had Claude collect feature requests spread across Slack. In about 26 minutes it found roughly 24 accounts, with each requester's handle, team, and a link to the original ask.
- Sorting a week of customer issues: in a second job, Claude condensed about 120 raw findings into 23 open issues and 14 resolved ones in roughly 50 minutes, work the requester estimated at "at least a week of full-time work."
- Speeding up legal review: Anthropic's product counsel runs marketing collateral through Claude in a dedicated channel, which cut review turnaround from a day or longer to about 30 minutes per asset.
Anthropic's docs also list the jobs it's built around, and they're a good map of where it fits:
- Watching monitors and alerts: scheduled dashboard checks and alert investigation, with a tool like Datadog, Sentry, or PagerDuty connected.
- Triaging requests: answering what it can in an intake channel, flagging duplicates, and routing the rest.
- Finding answers in your docs: policy and runbook questions answered with the source, through Google Drive, Notion, or Confluence.
- Answering data questions: turning a plain-language question into a warehouse query and a chart.
- Tracking projects: posting status digests and following up until approvals land.
- Fixing bugs: turning a bug report in the channel into a draft pull request on GitHub.
If the whole team should see the request, the progress, and the result, Claude Tag is a good fit; if it's your own research or drafting, Anthropic points you to Cowork or Claude Code.
Claude Tag pricing: does it cost extra?
Yes, Claude Tag costs extra on top of your Claude plan, because channel work is billed by usage. Adding Claude to Slack doesn't add a per-seat charge, but every task Claude does in a channel draws from a usage balance your organization funds.
So there are two pieces to the bill: the Claude plan you need to get access, and the usage Claude Tag burns through once it's running.
What you pay for
You need a Claude Team or Enterprise plan first. These are the list prices on claude.com/pricing as of October 11, 2026:
Anthropic doesn't publish a per-task cost for Claude Tag, and its spend-limit docs say so plainly: there's "no published per-task cost guidance."
The spend limit counts usage at list price (negotiated discounts apply on the invoice), and the model you pick plays a big part in what each task costs (fast mode costs more per token, for example).
Slack users don't each need a Claude seat to tag @Claude in a channel, so a 40-person channel doesn't mean 40 Claude seats.
What's billed and what isn't
Here's how Anthropic's spend-limit docs split it:
Claude can read a busy channel all day without costing anything: the meter only starts when it opens a working session, whether someone tagged it or it picked up work on its own.
Spend limits and alerts
A Primary Owner or Owner sets an organization-wide spend limit, which is a hard cap across every channel. If setup doesn't include the credit-purchase step, the launch step lets you pick $500, $1,000, $2,500, $5,000, Unlimited, or a custom amount up to $1,000,000 a month, with $2,500 preselected. The limit resets at the start of each month.
You can also set per-channel limits on top of the org cap, and new channels inherit a default. Alerts start at 75% of any limit, and Claude also posts a heads-up in the channel it's working in, so everyone there knows the budget is running low.
The Help Center also says work that would go over a limit "is declined, never silently cut short," so the person who asked can request more from an admin without leaving Slack.
How you fund the balance depends on your plan:
- Team plans: you have to buy usage credits before Claude responds in any channel.
- Invoiced Enterprise plans: usage bills to your invoice with no upper bound until you set a limit, so set one before your pilot starts.
The launch credit has ended
At launch, Anthropic gave eligible organizations a one-time Claude Tag credit: $25,000 per Enterprise organization and $2,500 per Team organization, covering channel usage only. The promo was extended once, and from September 2 organizations had to connect Claude Tag to GitHub to receive it.
According to the promo terms, all of those credits expired at 11:59 PM Pacific Time on October 1, 2026. If you set Claude Tag up now, every task draws from your own balance from day one.
How to estimate your bill
Since there's no rate card for Claude Tag tasks, Anthropic's own advice is to run a pilot. Set a spend limit you're comfortable with for the first month, use Claude in a couple of channels, and watch the per-channel breakdown on the usage page.
The analytics page also splits spend into Engaged (someone tagged or replied to Claude), Proactive (work Claude started on its own), and Scheduled (recurring jobs). If Proactive is eating your budget, you can turn off a channel's automatic replies and keep Claude to tagged requests.
Who can get Claude Tag?
Claude Tag is available on Claude Team and Enterprise plans, on Anthropic's own service. It isn't available on Free, Pro, or Max, or through third-party deployments. If you're still deciding which Claude plan you need at all, our Claude vs ChatGPT comparison covers the individual plans.
A few other things can block you even on the right plan:
- Data settings: organizations with Zero Data Retention, customer-managed encryption keys, or the HIPAA configuration for Claude Code and Cowork local mode can't use Claude Tag, because it stores channel memory and session transcripts.
- Routines: an admin has to turn on Routines under the organization's capabilities settings. Until then, Claude replies to every mention that it's unavailable.
- Roles: only a Primary Owner or Owner of your Claude organization can set it up. The Admin role can't.
Once it's running, an Owner decides who can use it. By default, anyone in the Slack workspace can use it. On Team you can limit it to members of your Claude organization, and on Enterprise to specific roles.
How to set up Claude Tag in Slack
Setup takes one Owner, one Slack workspace admin, and usually less time than it takes to get both of them online at once. Here's the flow from Anthropic's setup guide:
- Check the prerequisites: you need the Owner role in your Claude organization, a Slack workspace admin for the pairing step, Routines turned on, and (on a Team plan) usage credits.
- Start setup: go to claude.ai/admin-settings/claude-tag, click Start setup, then add the Claude app to Slack from the Slack Marketplace.
- Pair your workspace: in Slack, invite Claude to any channel with /invite @Claude, then have a Slack workspace admin send @Claude connect with no other text. Paste the pairing code on the setup page within 15 minutes (it expires) and click Pair workspace.
- Connect GitHub (optional): if your team will give Claude code work, a GitHub organization owner installs the Claude GitHub App and grants repositories. You can skip this and do it later.
- Fund usage and set a spend limit: if your organization pays by card in US dollars and has no usage credits yet (Team or self-serve Enterprise), setup asks you to buy some here (you can skip it), and you set the spend limit right after launch. Otherwise, you pick the monthly limit on the launch step.
- Add Claude to channels and launch: select at least one public channel and click Launch Claude Tag. To add Claude to a private channel later, run /invite @Claude inside it.
- Give Claude its own tool accounts: for each tool other than GitHub, create an email address for Claude, add it to the tool as a member with the narrowest role that covers the work, create an API key in Claude's account, and add that key as a connector in the Claude Tag admin page.
- Test it: in a channel Claude joined, send @Claude summarize what this channel decided this week and list any open questions. A working setup replies in a thread, with a footer naming the model. Then ask @Claude what can you access from this channel? to confirm your connectors show up.
Before people start tagging, check two settings. The guest setting defaults to Restrict, which blocks Claude from any channel with Slack guests (you can switch it to Channel only or Full access).
The "Let members know they can now tag Claude" toggle sends every workspace member a DM at launch, which is great for adoption and less great if you meant to keep the pilot small.
Moving over from Claude in Slack
Claude in Slack switched to Claude Tag on August 3, 2026, per the Help Center. If your team used the old app, the same setup steps apply and your existing app stays installed, but a Primary Owner or Owner has to opt the organization in.
Anthropic describes Claude Tag as the next generation of that app: it remembers context across days, schedules its own follow-ups, checks in on its own, and acts under its own identity.
{{cta}}
Claude Tag vs Cowork vs Claude Code
Anthropic has three products that can all do real work with your tools, and the docs draw a simple line between them: team work goes to Claude Tag, personal work goes to Cowork or Claude Code.
If you live in Slack and your team hands work back and forth all day, Claude Tag is the one built for that. If you mostly want Claude's help on your own files, you don't need it.
Claude Tag vs Lindy
We build Lindy, an AI teammate that also lives in Slack, so we're not a neutral party in this comparison. To keep it fair, every detail below comes from each company's own pricing page and docs as of October 11, 2026.
Both put an AI teammate in your Slack channels that anyone can tag with a task and follow in the thread. They split on who pays, how far the teammate reaches past the channel, and who's covered for regulated data:
Who pays, and how much
Claude Tag doesn't charge per person, so a 40-person channel can lean on it without 40 new seats. The catch is that the bill depends on what Claude does, and Anthropic doesn't publish a per-task cost, so you find your number by running a pilot under a spend limit.
Lindy charges per person. Everyone who uses it takes a seat at $49.99 a month, each seat adds 5,000 credits to a pool the whole team shares, and new teammates who join through Slack get 7 days free before their seat is billed. Lindy also publishes rough credit ranges: everyday asks like answers, summaries, and drafted replies use 2 to 250 credits, and deep work like a competitor report uses 250 to 1,000. When the pool runs low, Lindy pauses and tells you, and admins can top up at $10 per 1,000 credits.
Which one costs less comes down to how many people use it. A big channel where a handful of people hand Claude the occasional job suits Claude Tag's usage billing, while a team that leans on its AI teammate every day gets a bill it can predict from the seat count with Lindy.
What each one does outside the channel
Claude Tag is built around shared channel work, and Anthropic's docs send personal research and drafting to Cowork or Claude Code. Inside that lane, it's strong at engineering work: bug reports turn into draft pull requests through the Claude GitHub App, and tasks run in the same sandbox as Claude Code on the web.
Lindy covers the channel and each person's own workload. It triages your inbox and drafts replies, records and summarizes calls on Google Meet, Zoom, and Teams, preps you before meetings, sends a daily brief, and takes tasks by text over iMessage or SMS. It connects to 1,000+ tools plus any MCP server, and you choose per tool whether it writes on its own, asks for approval in the Slack thread first, or never writes at all.
Which one should you pick?
Pick Claude Tag if you:
- Already pay for Claude Team or Enterprise: you only add usage, and nobody in Slack needs a new seat.
- Hand it a lot of code work: draft pull requests from bug reports are what the GitHub App and Claude Code sandbox are built for.
- Expect lots of occasional users: usage billing suits a big channel where most people tag it now and then.
Pick Lindy if you:
- Want one teammate for channel work and personal work: the same Lindy handles your inbox, meetings, and calendar.
- Handle health data: Lindy's Enterprise plan comes with a signed BAA, and Claude Tag isn't covered by Anthropic's.
- Want a price you can plan around: seats set the bill, and credit ranges are published by task size.
- Aren't on Claude Team or Enterprise: Lindy works with free and paid Slack plans and needs no other subscription.
- Want to choose the model: pick from most major models for each task.
Claude Tag limits worth knowing
These are the constraints most likely to trip up a Claude Tag rollout:
- It's still in public beta: Anthropic's docs warn that features and behavior may change before general availability.
- It only works in Slack: Microsoft Teams support exists as a waitlist page that says Claude Tag "is coming to Microsoft Teams," with no date attached.
- There's no cost guidance: you can cap spend, but you can't predict it from a rate card, which makes budgeting a pilot exercise.
- Setup needs two specific people: a Claude Owner and a Slack workspace admin, and the pairing code expires after 15 minutes.
- Guest channels are off by default: shared channels with clients or contractors need a settings change, and Claude runs with reduced access while a guest is present if you pick Channel only.
- Sandbox files don't stick around: anything Claude made only inside the sandbox disappears when the sandbox is released, so for long tasks, ask it to push branches and post drafts as it goes.
- Deleting a Slack message doesn't delete it from Claude: everything Claude read stays in the session's transcript on Anthropic's side, even if you delete the message or thread in Slack.
Is Claude Tag worth it?
For a team that's already on Claude Team or Enterprise and runs its work through Slack, Claude Tag is one of the easiest ways to get Claude doing shared work, mostly because everyone sees the same thread and the same result.
The Anthropic examples (an estimated week of issue sorting done in about 50 minutes, legal review cut to 30 minutes per asset) show what it looks like when it fits.
Claude Tag is a good fit if you:
- Already pay for Claude Team or Enterprise: the setup cost is mostly an afternoon and a spend limit.
- Hand work around in Slack all day: shared threads, standing digests, and intake channels are where it shines.
- Have engineers in the loop: bug reports that turn into draft pull requests are one of its strongest use cases.
- Can live with usage-based billing: you'll need to watch spend for the first month or two.
Skip Claude Tag (for now) if you:
- Are on Free, Pro, or Max: you'd need to move your organization to a Team or Enterprise plan first.
- Run your company on Microsoft Teams: join the waitlist and check back.
- Need a fixed, predictable price: usage billing with no per-task guidance is a tough sell to a finance team that wants a number.
- Have ZDR or HIPAA requirements: those configurations block Claude Tag entirely.
If you're comparing options, our explainers on SpaceXSI's Grok Bot and OpenAI's ChatGPT dots cover two other AI teammates getting attention right now, and our roundup of Claude alternatives covers the broader field.
If all you need is a bot that answers a few questions in one channel, you can build a simple Slack bot without any of this.
Start Claude Tag with a small pilot
If you're already on Claude Team or Enterprise, pick one busy channel, set a spend limit you're comfortable with, and check the usage page's Engaged, Proactive, and Scheduled split every week for a month. If Proactive spend climbs faster than you expected, turn off that channel's automatic replies before you roll Claude Tag out wider.
Frequently asked questions
Does Claude Tag cost extra?
Claude Tag costs extra on top of a Claude Team or Enterprise plan. Channel work bills by usage to your organization's usage balance, capped by a spend limit an Owner sets. One-to-one DMs from someone with a connected Claude account bill to that person's own seat, and Claude reading a channel isn't billed.
Is Claude Tag available on Claude Pro or Max?
Claude Tag isn't available on Claude Pro, Max, or Free. It's limited to Claude Team and Enterprise plans on Anthropic's own service. Slack users in a paired workspace don't each need a Claude seat to tag @Claude in a channel.
Does Claude Tag work in Microsoft Teams?
Claude Tag doesn't work in Microsoft Teams yet. It works in Slack only. Anthropic runs a Microsoft Teams waitlist that says Claude Tag is coming to Teams, but it hasn't given a date.
What model does Claude Tag use?
Claude Tag launched on Opus 4.8, and Owners now choose its model. They pick the model per organization, workspace, or channel from the Opus and Sonnet models their organization allows. Channel members can also switch a thread to another model by asking Claude, unless an admin has blocked channel edits. Anthropic's Help Center lists Claude Tag among the places you can use Claude Opus 5.5, and each reply's footer names its model.
Can Claude Tag read private channels?
Yes, Claude Tag can read a private channel once someone invites it, for example with /invite @Claude. Anthropic says it doesn't report from private channels, and what it learns there stays in that channel's own notes. From a private channel, Claude can read public channels it's in but can't post to them.
How do I enable Claude Tag?
A Primary Owner or Owner enables Claude Tag at claude.ai/admin-settings/claude-tag. They pair the Slack workspace with a code from @Claude connect, fund usage and set a spend limit, add Claude to at least one public channel, and click Launch Claude Tag.
