Open-source AI agent harness 'QM,' released by Y Combinator, is a system that helps teams collaborate with AI agents to handle practical tasks like managing emails and repository maintenance.
Imagine this: you wake up in the morning and open your email inbox, only to find that the dozens of inquiries that came in last night have already been categorized by priority, with draft responses already written. Or, during a team project, you drop a comment in Slack like, “Please reflect the action items from last week’s meeting minutes into the repository,” and the actual coding work starts automatically.
Until now, AI has been a smart conversational partner that answers our questions. However, we are moving into an era where AI doesn’t just talk, but performs actual “work” as a member of the team. Recently, Y Combinator (YC), known as the cradle for startups, pushed this future closer by open-sourcing ‘QM,’ an AI collaboration system it has been using internally. Source: Y Combinator on X, Source: QM — Open-Source Agent Harness from YC
Why is this important?
Many AI tools we have encountered so far have focused on increasing the productivity of ‘individuals.’ But real-world work usually happens at a ‘team’ level. Some tasks require the permissions of the accounting team, and others require code from the engineering team.
QM combines this team-based collaborative environment with AI. It goes beyond the AI serving as a personal assistant, enabling an entire enterprise to work with AI agents in a single, massive ‘multiplayer’ environment. Source: YC QM Agent Harness: A Collaborative AI Shift, Source: QM: Multiplayer AI Agent Harness for Startups and Slack YC associates agree that this tool allowed them to work as efficiently as an army, even with far fewer personnel. Source: eve on X
Understanding it Simply: AI’s ‘Custom Work Uniform’
The term ‘Agent Harness’ might be unfamiliar. Simply put, think of the AI model as the ‘brain,’ and the harness as the ‘custom work uniform’ that enables that brain to communicate with the world and perform practical work.
| An agent harness is software that wraps around an AI model. [Source: What Is an Agent Harness? Model vs Agent(2026) | Taskade AI](https://www.taskade.com/wiki/ai-agents/agent-harness) It grants the AI—which was previously only capable of predicting text—the authority to create work plans, read and write files, and use external tools. |
To use an analogy, it’s like a very smart college student (AI model) who knows how to read documents but can’t get any work done because they lack a company intranet ID or the official form templates (the harness). The harness gives this student an ID, an operations manual, and a company stamp. QM is a ‘multiplayer harness’ designed so that the entire team can share and wear this work uniform. Source: QM — Open-Source Agent Harness from YC, Source: Agentharness: что это, компоненты и примеры (2026)
Current Status and Features
QM has been carefully designed to be used directly in the corporate workplace.
- Harmony between Individual and Team: It allows for individual personalized settings while maintaining a work environment shared by the whole team. Source: YC QM Agent Harness: A Collaborative AI Shift
- Security and Audit: This is the most important part. AI agents act as delegates, using the credentials (IDs, permissions, etc.) of the person who assigned the work. Furthermore, since every action performed by the AI is logged, it can be managed transparently, making it secure. Source: GitHub - yc-software/qm
- Flexibility: You can issue work instructions through direct conversation via Slack or web interfaces, and administrators can set which models to use and the required security levels according to the organization’s needs. Source: YC QM Agent Harness: A Collaborative AI Shift, Source: QM — Open-Source Agent Harness from YC
What Lies Ahead?
QM has been released as open-source under the MIT License. This means that developers around the world can customize and further improve the system YC built based on their own situations. Source: Y Combinator on X Integrations with various collaborative tools used by enterprises beyond Slack are expected to increase rapidly.
AI is evolving from an entity that just answers questions when asked into a ‘digital colleague’ that performs actual work and collaborates with teammates. Your team might soon have a digital colleague like QM join its ranks.
References
- GitHub - yc-software/qm: Multi-player agent harness for work · GitHub
-
[What Is an Agent Harness? Model vs Agent(2026) Taskade AI](https://www.taskade.com/wiki/ai-agents/agent-harness) - Agentharness: что это, компоненты и примеры (2026)
- Y Combinator on X
-
[YC QM Agent Harness: A Collaborative AI Shift StartupHub.ai](https://www.startuphub.ai/ai-news/artificial-intelligence/2026/yc-qm-agent-harness-a-collaborative-ai-shift) - QM — Open-Source Agent Harness from YC
- eve on X
- QM — Open-Source Agent Harness from YC
- QM: Multiplayer AI Agent Harness for Startups and Slack
- QM: A Multiplayer Agent Harness Built for Secure Team Workflows
- Personal game play assistance
- Automating and managing collaborative tasks at a team level
- Developing AI models themselves
- A term referring to the AI model's brain
- A software shell that enables an AI model to perform actual work
- Physical components of a computer
- No security, everyone has access to all data
- It uses the user's credentials as a delegate, and all actions are audited
- Only administrators perform all tasks