What they are doing
QM is an open-source project published under yc-software, the GitHub organization identifying itself as Y Combinator. It provides the software around an AI agent so several employees can work with it through Slack and a web interface. It is a project to deploy and operate, with its own contributor community. Publisher, project overview.
Its central idea is a “scope”: a personal or shared workspace with its own files, memory, permissions and working computer. Employees can use individual spaces and collaborate in channels or projects. Administrators select available models and agent software, manage sharing, and control which reusable instructions become available across the organization. QM overview.
Latest project and updates
The latest verified release is QM 0.1.12, dated September 19, 2026 on GitHub. It introduces agent swarms in beta: groups of agents with persistent work histories, scoped messages and visible progress. Other released changes include provider-managed Slack installation, email archiving and a Calendar tab, improved management of internal apps, and connectors that can use individual employees’ credentials. These are release-note claims, not results from our testing. Version 0.1.12.
The swarm documentation explains that workers receive separate computers and do not automatically receive the initiating agent’s files or approval grants. The feature therefore requires deliberate work assignment and access configuration; creating more agents does not automatically give them the information needed to finish a task. Agent swarms.
Why it matters to your business
Our interpretation: QM addresses a practical problem that appears when several people use AI: keeping shared work useful without treating every employee’s information as common property.
Two possible applications are a weekly project summary assembled from an approved team workspace, and reusable instructions for preparing consistent internal handoffs. A third is a small internal app for tracking an operational task. These suggestions build on QM’s documented shared skills, scheduled work and internal-app capabilities; they are not promises of time savings. Documented capabilities.
Picture this
Imagine a six-person property-maintenance business. Its coordinator prepares tomorrow’s job handoffs from an approved work-order list, access instructions and completion checklist. Crew leads review the summaries in a shared space. The owner keeps employee compensation and private correspondence outside that task.
The first success measure would be straightforward: whether a crew lead can identify the correct job, required materials, responsible person and completion evidence without another phone call. This is an illustrative workflow, not a customer case study. Any deployed version would still need actual access checks: QM’s policy explicitly says its isolation design does not guarantee that information cannot leak. Security policy.
Where they are heading
The current development direction is visible in the beta swarms and the expansion of collaboration, mail and app management. Our interpretation is that QM is becoming a broader environment for ongoing team work. The release establishes what has shipped, while the swarm documentation explicitly defers a dedicated agent-board interface. No delivery date for that interface was verified. Latest release, swarm documentation.
Try the idea
Choose one recurring handoff. Write down the task, approved inputs, recipient, owner and definition of done. Use your existing AI application to prepare a draft from non-sensitive example information. Have the recipient check it against the source records. Establish whether that workflow helps before adding shared agents or automation.
When the product itself may fit
QM may fit a technically supported team that needs persistent shared AI work. It is a poor fit for an owner seeking a maintenance-free application or a ready-made public customer chatbot. The project calls itself early, experimental software and says its interactive agent surfaces assume authenticated internal users from one organization. Security limitations.
Setup involves deployment configuration, model access, a database, credentials and a supported runtime environment. The deployment documentation covers Docker, Fly and AWS, with different prerequisites. Software licensing and operating costs are separate: QM carries an MIT license, while hosting, model usage, connected services and maintenance need their own budget. A dependable total monthly price was not verified. Deployment requirements, license.
Explore the original
Start with the official QM repository and latest verified release.
Sources checked September 20, 2026. Evidence: public documentation reviewed; no hands-on product trial.
Sources checked September 20, 2026. Documentation and publisher announcements reviewed; no hands-on product trial. Business examples and resource matches are our interpretation.