There is no native Claude app for Microsoft Teams in October 2026, so the best way to use Claude in Teams depends on which of four things you want: a bot in your channels, Claude inside Microsoft’s Copilot, Claude reading Teams from the outside, or a workflow you build yourself. PlugAnd.ai is the only turnkey option we found that puts Claude in a shared channel without per-user seats.

The question has changed in the past year. Microsoft has made Anthropic’s models part of its own stack, and Anthropic’s connector now reaches into Teams, but neither is the same as typing a question to Claude in a channel thread. Most roundups blur these together.

Four meanings of “Claude in Teams”

Claude as a bot in the channel. A colleague @-mentions an assistant, it answers in the thread, and everyone sees the reply. This is what most people picture, and today it requires a third-party app or something you build.

Claude inside Microsoft’s products. Microsoft Learn says Anthropic models are available in Microsoft Copilot, Researcher, Copilot Studio, Power Platform and Copilot in Microsoft 365 apps. You use Microsoft’s interface, Microsoft picks where Claude appears, and the features that read your Teams data sit behind a paid license.

Claude reaching into Teams. Claude’s own connector searches your chats and channels from inside Claude, and with write tools enabled can post. The work happens in Claude, not in Teams.

Claude triggered by a workflow. A message arrives, a scenario calls the Claude API, a reply is posted. Useful for one job, not a conversation.

The model matters more than the seat

PlugAnd.ai’s pricing page includes a worked example for a ten-person tenant: 4,200 requests a month at 3,000 input and 700 output tokens each. On a budget GPT model (listed at $0.20 input and $1.20 output per million tokens) that comes to about $6.72 of top-up; on Claude Sonnet 5.5 ($2.00 and $10.00) about $60.67. We re-ran the arithmetic and it holds, including the 10% top-up fee. The vendor labels it an illustration, not a bill, and long threads, attachments, reasoning tokens and web search all add cost.

On a usage-billed assistant, Claude should be a deliberate choice for the questions that need it, not the default for every message. A per-seat license hides that trade-off; a shared balance makes it visible.

Data path and region: what to check first

If you route through Microsoft, the data terms are Microsoft’s: Learn says its Data Protection Addendum applies to Anthropic models used through Microsoft’s Online Services, except models labeled “Anthropic models with Data Retention,” which have to be separately enabled and fall under Anthropic’s own terms. Learn also says Anthropic models are currently excluded from the EU Data Boundary and are off by default in the EU, EFTA and UK. European tenants should settle this first. It applies to Microsoft’s route, not to a third-party bot, which has its own terms to review.

Who should pick what

A team that wants to ask Claude in channels, with no seats: PlugAnd.ai, using a cheaper model for routine questions.

A company that needs Claude on its own chats, files and meetings: Microsoft 365 Copilot, with Anthropic enabled by an admin.

An IT team building governed, domain-specific assistants: Copilot Studio agents with Claude selected as the model.

Individuals who already pay for Claude: the Microsoft 365 connector, with write tools left on Ask or Blocked until you trust the workflow.

One repeatable task: a Make, Zapier or Power Automate scenario.