Hermes AI agent arrives in Microsoft Teams

Nous Research has brought its open-source Hermes Agent into the Microsoft tools that offices already use, announcing on 18 November that the agent now works inside Microsoft Teams, the Microsoft Store and Microsoft Azure. The move, unveiled during Microsoft Ignite in San Francisco, turns a popular open agent into an enterprise product aimed squarely at everyday office work.
Nous, a portfolio company of M12, Microsoft's venture fund, says Hermes is the top-ranked public open agent on OpenRouter, the marketplace where developers compare AI models, and that it processes nearly two trillion tokens a day. Both figures come from Nous itself and have not been independently verified.
Teams, the Store and Azure
Hermes Enterprise is built to sit inside the Microsoft workplace in three places. Hermes Desktop for Windows is available in the Microsoft Store. In Microsoft Teams, employees can hand the agent knowledge work, from financial analysis and drafting documents to triaging an inbox, and review the finished result in the same conversation. The agent workloads themselves run on Azure Container Apps Sandboxes, isolated microVMs that start in under a second and release computing power when idle.
The pitch is that staff should not have to leave the apps they already use to get an AI agent to do a job. Nous chief executive Dillon Rolnick said the agent "gets better the more you use it", and that the Microsoft integration means "what one person figures out becomes shared intelligence the whole company can build on".
Collective Wisdom turns tricks into team skills
The centrepiece of the launch is a feature called Collective Wisdom. As an employee delegates recurring work, Hermes automatically turns a successful approach into a reusable skill: a written definition paired with code that runs the stable steps the same way every time, so established workflows use less inference and become more reliable. Employees can then ask Hermes which of their skills are worth sharing, and propose and publish them for their team, with review.
The idea is that the next person to attempt a task starts from the best approach anyone on the team has found, rather than from scratch. Nous says the shared knowledge stays a private asset of the organisation and does not become training data for an AI vendor. The announcement does not spell out who reviews a proposed skill or what that review involves, a gap that teams considering the feature will want clarified before they switch it on.
An option to run it in your own tenant
For organisations with strict data residency, security or procurement requirements, Hermes Enterprise can run in the customer's own Azure tenant under a similar product and pricing model, Nous said. Michelle Gonzalez, global head of M12, said the partnership gives organisations "a simple way to put self-improving agents to work and build shared intelligence across their teams".
Nous presented the work at Microsoft Ignite, which ran from 17 to 20 November, in session BRK448, "Run secure agent loops in production with isolated sandboxes at scale". The launch puts an open-source agent inside the world's most widely used office software, at a moment when every major AI lab is racing to move its agents from chat windows into the places where work actually happens.