Back to list
QM: A New Multiplayer AI Agent Harness for Collaborative Startup Workflows in Slack and Web
Product LaunchAI AgentsOpen SourceStartup Tools

QM: A New Multiplayer AI Agent Harness for Collaborative Startup Workflows in Slack and Web

QM is an innovative multiplayer agent harness designed specifically for the startup environment, bridging the gap between personal AI assistants and company-wide automation. Operating across Slack and web interfaces, QM provides isolated workspaces for individual employees while enabling seamless collaboration in shared channels and projects. The platform features scoped memory, granular permissions, and a durable sandbox for each user and room. Notably, QM is vendor-agnostic, allowing users to switch between models like Pi, OpenCode, Codex, and Claude Code. Its capabilities range from searching internal databases and triaging emails to managing GitHub repositories and deploying custom internal web apps, all powered by a headless core and Postgres-backed architecture. This tool aims to streamline startup operations by providing a flexible, secure, and highly customizable AI infrastructure.

Hacker News

Key Takeaways

  • Multiplayer Design: Unlike traditional personal AI assistants, QM is built for startups, allowing employees to have isolated workspaces while collaborating in shared Slack channels and projects.
  • Vendor Agnostic: The platform is built with an open-source mindset, supporting various models and harnesses including Pi, OpenCode, Codex, and Claude Code without vendor lock-in.
  • Comprehensive Scoping: Each person and room maintains its own scoped memory, files, keychain view, permissions, and durable sandbox for secure and organized operations.
  • Advanced Automation: QM supports background work through crons and watches, enabling tasks to run autonomously, such as monitoring CI or triaging inboxes.
  • Customizable Infrastructure: Users can build and publish internal web apps, manage shared skills via git repositories, and set organization-level security postures.

In-Depth Analysis

Redefining AI Agents for the Startup Ecosystem

The current landscape of AI agents is largely dominated by personal assistants designed for individual use. While these tools are effective for personal productivity, scaling them to support an entire organization often leads to significant complexity and management overhead. QM addresses this gap by introducing a "multiplayer" harness specifically tailored for startups. The core philosophy of QM is to provide a dual-layered environment: individual isolation and collective collaboration.

In this framework, every employee receives an isolated workspace. This ensures that personal tasks, configurations, and data remain private and do not interfere with the work of others. However, the true power of QM lies in its collaborative capabilities. The agent can be integrated into Slack channels, group messages, and specific projects, allowing teams to interact with the AI as a shared resource. This structure is supported by a robust scoping system where memory, files, and permissions are tied to specific users or rooms, ensuring that the AI context is always relevant and secure.

Technical Flexibility and Open Source Integration

One of the standout features of QM is its commitment to a vendor-agnostic architecture. In an era where AI models are rapidly evolving, being tied to a single provider can be a strategic risk. QM mitigates this by allowing users to pick their own harness and model. Whether a startup prefers Pi, OpenCode, Codex, or Claude Code, the core engine remains the same. This flexibility ensures that a deployment can adapt to the best available technology at any given time.

From an architectural standpoint, QM utilizes a headless core supported by a Postgres database for managing sessions, memory, and task queues. This backend allows for sophisticated features such as "durable sandboxes" and "keychain views," which are essential for handling sensitive credentials and executing code safely. Furthermore, the platform extends its utility through "Shared Skills." These skills are scope-owned and can be shared across the organization via administrative grants or imported as skill packs from git repositories, creating a modular and scalable AI capability set.

Operational Impact: From Repo Management to Email Triage

QM is not just a conversational interface; it is a functional tool designed to perform heavy-lifting tasks. For developers, QM can work directly within existing repositories to run tests, open pull requests, monitor continuous integration (CI) pipelines, and check system logs. This level of integration allows the agent to act as a junior engineer or a DevOps assistant, tracking project progress in shared channels and posting automated updates.

For administrative and operational roles, QM offers powerful data retrieval and communication tools. It can search across internal notes, emails, documents, and databases simultaneously to act as a "company brain." One of its most specialized features is the ability to learn a user's writing voice from past communications. This allows the agent to triage inboxes on a schedule, creating labels and drafting replies that maintain the user's professional tone. Additionally, the ability to spin up custom internal web apps and publish them to specific groups within the company allows startups to build bespoke internal tools quickly without traditional development cycles.

Industry Impact

The introduction of QM signals a shift in the AI industry toward "agentic workflows" that are deeply integrated into existing communication platforms like Slack. By focusing on a multiplayer harness, QM highlights the growing need for AI tools that understand organizational hierarchy and collaborative dynamics rather than just individual prompts.

Furthermore, the emphasis on open-source compatibility and model flexibility challenges the dominance of closed-ecosystem AI providers. As startups look for ways to integrate AI without compromising data sovereignty or becoming overly dependent on a single vendor, platforms like QM provide a blueprint for a more modular and secure corporate AI infrastructure. The inclusion of background automation (crons) and durable sandboxes also points toward a future where AI agents are not just reactive participants but proactive members of a technical team.

Frequently Asked Questions

Question: How does QM handle data privacy between different team members?

QM uses a system of personal and shared scopes. Each employee has an isolated workspace with its own scoped memory, files, and permissions. This ensures that an individual's private work and data are not accessible to others unless explicitly shared in a collaborative channel or project room.

Question: Can I use QM with my preferred AI model?

Yes. QM is designed to be vendor-agnostic. It supports multiple harnesses and models, including Pi, OpenCode, Codex, and Claude Code. Users can switch between these models while maintaining the same core deployment and configuration.

Question: What kind of automated tasks can QM perform?

QM can handle a variety of background tasks using crons and watches. This includes triaging emails and drafting replies on a schedule, monitoring CI/CD pipelines, running tests in repositories, and searching across multiple internal data sources like databases and documents to retrieve information.

Related News

Gemini API Tool Retrieval Preview Gives Agents Access to Hundreds of Tools With Fewer Tokens
Product Launch

Gemini API Tool Retrieval Preview Gives Agents Access to Hundreds of Tools With Fewer Tokens

Google AI Blog announced that Tool Retrieval is now in preview in the Gemini API. The feature is presented as a way to give agents access to hundreds of tools while using fewer tokens. The announcement pairs those two points: the breadth of tool access and lower token use. It does not explain how Tool Retrieval works, specify how token use was measured, or quantify the reduction. It also does not provide an exact tool count beyond describing access to hundreds of tools. The stated release status is preview, and the announcement does not give a timeline for general availability.

LangSmith Tool Gateway Enters Public Beta for Secure, Governed AI Agent Tool Use
Product Launch

LangSmith Tool Gateway Enters Public Beta for Secure, Governed AI Agent Tool Use

LangChain announced that LangSmith Tool Gateway is in Public Beta. The gateway handles credentials, identity, and access for every agent tool call, with policies, approvals, and traces. The announcement frames secure, governed tool use around the individual call: its focus is on how agents access tools, rather than on what those tools or agents can do. The update identifies what the gateway covers but does not explain how credentials are managed, how access decisions are made, when approvals apply, or what information traces contain. The Public Beta is the only release status stated in the announcement.

Google’s Gemini Decision 1 Picks Answers Instead of Writing Them Across Five Input Types
Product Launch

Google’s Gemini Decision 1 Picks Answers Instead of Writing Them Across Five Input Types

Google’s Gemini Decision 1 is a model designed to pick answers rather than write them, according to Shittu Olumide’s report for KDnuggets. It reads text, images, video, audio, and PDFs in a single forward pass and supports up to 262,144 input tokens per request. The central distinction is between what the model reads and what it does with that input: it accepts several kinds of material, but its stated task is answer selection, not answer writing. The token figure describes the maximum input per request, not how much the model outputs. The report does not explain how answers are presented for selection or provide results showing how well the model performs.