Quorum: A Slack Decision Memory Agent Based on RTS + MCP + Vercel Workflow

Quorum is an AI decision memory agent designed specifically for Slack, combining Real-Time Sync (RTS), Model Context Protocol (MCP), and Vercel Workflow to address the problem of scattered decision records that are difficult to trace and reuse in team collaboration.

Original Author/Maintainer: OrionArchitekton Source Platform: GitHub Project Link: https://github.com/OrionArchitekton/quorum-slack-agent Release Time: June 8, 2026

The project focuses on capturing, organizing, and retrieving team decision information to serve as a knowledge assistant, and it participated in the Slack Agent Builder Challenge to demonstrate the deep integration of AI and collaboration tools.

In modern remote work environments, team decisions are often scattered across message threads, meeting minutes, and documents, making them hard to trace and reuse. Quorum was created to address this pain point by intelligently managing decision memory.

The project participated in the Slack Agent Builder Challenge to explore the deep integration of AI technology and enterprise collaboration tools to improve work efficiency.

Real-Time Sync (RTS)

Ensure the agent timely perceives Slack conversation dynamics, monitors specific channels/threads, identifies decision patterns, and automatically triggers recording when a decision is reached, keeping in sync with Slack's state.

Model Context Protocol (MCP)

An open standard proposed by Anthropic that supports seamless integration with external services (e.g., knowledge bases, Jira, Linear, calendar systems), expanding capability boundaries with strong scalability.

Vercel Workflow

Provides a reliable backend execution environment to handle asynchronous tasks, persistent storage, complex analysis (extracting decision points, identifying participants), scheduled tasks (decision review reminders), and third-party API interactions. The serverless architecture enables on-demand scaling.

Decision Capture

Intelligently identifies decision moments, records decision content, participants, key arguments, time, and context in a structured way, preserving the 'why' of decisions rather than just the 'what'.

Storage and Organization

A structured knowledge base supports multi-dimensional classification: indexed by project/team, decision type (technical/product/operational), timeline, and relevant personnel for easy browsing and retrieval.

Intelligent Retrieval and Recommendation

Proactively provides historical decision references: similar scenario options, relevant personnel positions, execution result feedback, avoiding repeated discussions and maintaining decision consistency.

Review and Tracking

Sets decision review reminders, records execution feedback, and forms a closed loop to continuously improve decision quality.

Monorepo Architecture

Managed using pnpm workspace, including apps/web (web application and API), packages (shared libraries), workflows (Vercel Workflow definitions), docs (documentation), with clear code organization.

Type Safety

Uses TypeScript strict mode to catch potential errors at compile time and reduce runtime failures.

Test Coverage

Configures Vitest to support unit, integration, and Workflow tests, ensuring reliable agent behavior.

Development Experience

Includes the .claude/hooks directory to support AI-assisted development tools and embrace modern AI workflows.

• Remote Team Decision-Making: Acts as a virtual decision recorder to solve the problem of distributed teams lacking face-to-face discussions.
• New Employee Onboarding: Quickly understand historical decision logic to accelerate integration.
• Project Retrospectives: Provides a complete decision timeline to support effective analysis.
• Knowledge Transfer: Preserves decision memory when team members change to ensure organizational knowledge continuity.
• Compliance Auditing: Automatically generates audit trails of decision processes.

• Protocol Standardization: The adoption of MCP reflects the move toward standardization in the agent ecosystem, where interoperability becomes a key competitive advantage.
• Platform Integration: Deep integration with Slack rather than being an independent application is the mainstream path for agent implementation.
• Serverless Architecture: Vercel Workflow reflects the migration of AI backends to serverless to handle uncertain loads.
• TypeScript Dominance: Type safety, toolchains, and the npm ecosystem make it the preferred language for AI development.

Quorum does not replace human decision-making; instead, it enhances decision-making capabilities through better memory, retrieval, and review. In an era of information overload, this 'decision memory' function is expected to become an important part of team knowledge management, bringing new efficiency improvements to enterprise collaboration.