Y Combinator Open-Sources QM Agent Harness

Y Combinator has open-sourced QM (quartermaster), a multiplayer agent harness designed for collaborative work that operates within Slack and on the web. The organization utilizes QM internally across various departments, including accounting, legal, events, and engineering, and even for building the QM project itself. This release is framed by Y Combinator as an early-stage experiment that may contain bugs, and it is made available under the MIT license.

QM is described as deployable today, functioning as organizational software rather than a standalone desktop application. Successful deployment necessitates a cloud account, a PostgreSQL database, and an individual comfortable with infrastructure management. The ideal users for QM are startups or mid-sized companies, typically ranging from 10 to 500 employees, with at least one dedicated platform engineer. Larger enterprises can implement QM but are advised to first review its security posture. Realistic industry applications include venture capital and professional services, fintech and accounting operations, legal operations, event management, and internal tooling for B2B SaaS companies.

Specific applications detailed within the QM repository include the ability to search across internal notes, emails, documents, databases, and the web simultaneously. It can also triage inboxes on a scheduled basis, applying labels and drafting replies. Furthermore, QM can interact with existing code repositories to execute tests, initiate pull requests, and monitor continuous integration processes. It also facilitates project tracking within shared communication channels. The deployment process for QM does not require cloning the repository directly. Instead, an organization-specific deployment repository depends on the `@yc-software/qm` package and then executes `qm init` with an organization slug and a specified target platform like Fly.io or AWS.

Upon initialization, QM materializes a deployment skill that guides users through setting up infrastructure, web sign-in, connectors, optional Slack integration, and live verification. The default sign-in mechanism employs a built-in broker that sends a one-time link via email, though an external identity provider can be integrated as an alternative. Each QM deployment operates within the cloud account of the respective operator, and the initialization process does not automatically create deployment CI pipelines. Teams seeking to maintain core QM functionality alongside their customizations within a single codebase are encouraged to create a private fork through a standard clone operation, rather than using GitHub's built-in fork feature. Further technical details are available in the `deployment.md` and `getting-started.md` files within the project's documentation.