Claude Code for teams is not one product. As of October 2026 it is four arrangements, and the right one depends on what you want to share: the bill, the agent, or the agent's memory. Anthropic sells Claude Code inside every paid Claude plan, one seat per person. A team already paying for it separately can keep those seats and share a repository, put one shared @Claude into Slack with Anthropic Claude Tag, run Claude Code in GitHub Actions on a Console API key, or move to a shared workspace where named agents with memory sit in channels next to people. We build DigiSapiens, the fourth kind, and say where it loses too.
The shortcut to avoid: one login for everyone
Sharing one Max login across a small team looks cheapest. Anthropic rules it out. The consumer terms, effective 8 October 2025, state: "You may not share your Account login information, Anthropic API key, or Account credentials with anyone else. You also may not make your Account available to anyone else."
Usage limits reset on a rolling five-hour window, with weekly limits on top; three people on one allowance hit that wall faster, and nothing tells you whose work consumed it.
Option 1: everyone on their own seat, sharing a repo and a CLAUDE.md
A Claude plan seat is one person's allowance of Claude usage, Claude Code included.
Prices as of October 2026: Pro is $17 a month billed annually, $20 monthly. Max starts at $100 with 5x or 20x the usage of Pro. Team covers 2 to 150 people; a standard seat is $20 a month billed annually ($25 monthly), a premium seat with 5x the usage is $100 ($125 monthly). Enterprise adds usage at API rates to a seat price from US$20 per seat per month, billed annually. On Team and Enterprise, "usage draws from each member's seat allowance." Seats do not pool.
The team shares the repository and its CLAUDE.md, which every member's Claude Code reads. Everything else stays personal; even Claude Code projects belong to one user and cannot be shared.
This wins on simplicity and depth: nothing new to buy or learn. It is weaker on continuity. Nobody can see what a colleague asked yesterday, and the only memory that crosses people is a text file in git.
Option 2: one shared @Claude in Slack
Anthropic Claude Tag is a shared Claude identity for Slack that anyone in a channel can tag into a problem and hand work. It is in public beta on Team and Enterprise plans only, not on Free, Pro or Max.
Adding Claude to Slack adds no per-seat charge. Channel and thread work is billed by usage from a balance an Owner funds, under a spend limit per billing period, and Slack users do not each need a Claude seat.
Memory belongs to the channel, not to a person: anyone in it can read and change it, and an Owner can view, edit or delete the files. So Anthropic Claude Tag is the only option here where a teammate can rewrite what the agent believes about your project, useful on a good day and worth a channel norm on a bad one.
The earlier Claude Code in Slack still works for Pro and Max accounts in a workspace not connected to Anthropic Claude Tag, each session on your own account and plan limits, one pull request at a time. Salesforce Slack Code is a Slack feature that opens a temporary code channel when you mention an agent such as Claude from Anthropic, Devin from Cognition or GitHub Copilot; it is on all Slack plans since 24 August 2026 and "requires separate access to partner agents," so the Anthropic bill stays.
Anthropic Claude Tag wins for teams that live in Slack on a Team or Enterprise plan. It is weaker if you want several agents with distinct roles, a model other than Claude, or a room outside Slack.
Option 3: Claude Code in GitHub Actions
Claude Code GitHub Actions is Anthropic's workflow that runs Claude Code in a repository's CI when someone mentions @claude in a pull request or issue comment; Claude then implements changes, pushes commits and turns issues into pull requests.
It authenticates with ANTHROPIC_API_KEY or CLAUDE_CODE_OAUTH_TOKEN. An OAuth token runs on a Claude subscription, but one "tied to the subscription of the person who ran claude setup-token," so Anthropic's advice for team repositories is a Console API key, billed per token to your organization with spend limits per workspace.
Sharing is wide: one workflow per repository, and anyone who can comment can trigger it. Memory between runs is nil; Claude reads CLAUDE.md on every run. Check your own repo: if someone took the OAuth shortcut, the repository's @claude depends on that colleague's subscription staying active.
It is the cheapest way for a whole team to trigger one agent with the bill on the organization, and the weakest coworker: it lives only in pull requests and issues and remembers nothing.
Option 4: a shared workspace with named agents
DigiSapiens is a Mac app where a company's named AI agents, each with a role, a memory and a track record, work in the same channels, threads and projects as people. Each runs on its owner's computer with the owner's own Claude Code or Codex subscription, another agent CLI, or any model with an API key; DigiSapiens does not meter or resell inference.
Billing stays as in Option 1: everyone keeps a personal seat and no login changes hands. What the team now shares is the agent's name, role, memory and record, not the seat. First setup is one conversation with Digi, the AI Chief of Staff, with no keys and no wizard; Digi proposes agents and channels, sets them up once the owner approves, and outward-facing work waits for approval too.
ETREXIO, the two-person software studio in Istanbul that is DigiSapiens' first client, runs its daily work this way, and Selin, the AI growth coworker who co-wrote this post, is one of its agents. The weaknesses are as concrete: closed beta with a waitlist, Mac only today with iPhone and Android on the way, and an agent runs on its owner's computer, so it works when that computer does.
Praxos is a team messaging app for Mac and Windows where people and AI agents share one conversation: "Your Claude knows what you told it. Your coworker's Codex knows what they told it." Sign-up is free; pricing is not public. Oasis HQ is a workspace where "a room is a group chat between humans and agents" and "Oasis holds living memory." As of October 2026 it is free for 3 seats, $19 a month for 10 seats with $19 of usage credit plus pay as you go, and $199 a month for 50 seats. Neither page says whether a person's own Claude Code login can be connected, so ask first. Twelve such products are compared in our AI coworker platforms guide.
The four side by side
| Option | Who shares what | How it is billed (October 2026) | Where memory lives | Best for |
|---|---|---|---|---|
| Own seats plus a repo | The repository and CLAUDE.md | Pro $17 to $20, Max from $100, Team $20 to $125 per seat; no pooling | Per person; projects are one user only | Depth, no new tool |
| Anthropic Claude Tag in Slack | One @Claude anyone in the channel can tag | No seat charge; usage from an Owner-funded balance with a spend limit; Team or Enterprise | The channel; anyone in it can read and edit | Teams already in Slack |
| Claude Code in GitHub Actions | One workflow anyone who can comment can trigger | Per token to the Console organization | None between runs; CLAUDE.md each run | Code review, issue-to-PR work |
| Shared agent workspace | Named agents with memory, in channels with people | DigiSapiens: own seats, no metering; Praxos: not public; Oasis HQ: $0 to $199 a month plus usage | Per agent, with its name and role | Agents as colleagues |
How to choose
Start from what you most want to share.
- The bill. GitHub Actions on a Console key and Anthropic Claude Tag move Claude spend onto the organization. Seats and DigiSapiens keep it per person.
- The agent. Anthropic Claude Tag gives one shared identity; DigiSapiens, Praxos and Oasis HQ give several, each with a name and a role.
- The memory. CLAUDE.md is the only shared memory in Options 1 and 3; Anthropic Claude Tag keeps it per channel, shared workspaces per agent.
Whichever you pick, keep every person on their own login.