How Slack Manages Context in Long-Running Multi-agent Systems

2026-04-28 · Source: InfoQ · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Advanced, quick

Summary

Slack engineers have developed a novel approach to managing context in long-running multi-agent systems, moving beyond simple chat log accumulation. This method, detailed in an April 28, 2026, engineering blog post, addresses the challenge of maintaining coherence and accuracy over hundreds of requests and megabytes of output, where full message history becomes impractical. Their system employs a "coordinator/dispatcher" multi-agent design, featuring a central coordinator, specialized experts, and critics. The core of their solution involves three complementary context channels: a director's journal for structured working memory, a critic's review for annotated findings with credibility scores, and a critic's timeline for chronological, credibility-weighted findings. This structured memory system allows agents to build on distilled truth, significantly improving the reliability of complex agentic applications.

Key takeaway

For AI Architects designing multi-agent systems that require sustained operation and accuracy, your approach to context management is critical. You should consider implementing structured memory channels, similar to Slack's director's journal and critic's review, to prevent context window limits and response degradation. This will enable your agents to maintain coherence and build on distilled, validated information over hundreds of interactions, rather than relying on accumulating chat logs.

Key insights

Structured memory, validation, and distilled truth are crucial for long-running multi-agent system coherence.

Principles

• Explicit context management is essential for long-running LLM sessions.
• Structured summaries maintain coherence across agent interactions.
• Credibility scoring identifies corroborated findings from multiple sources.

Method

Slack's method uses a coordinator/dispatcher design with a director's journal, critic's review (credibility-weighted findings), and critic's timeline (coherent narrative from credible evidence) to manage agent context.

In practice

• Implement a "director's journal" for agent working memory.
• Use "critics" to evaluate expert findings and assign credibility scores.
• Build a "critic's timeline" to resolve conflicts and retain credible evidence.

Topics

• Multi-agent Systems
• Context Management
• Structured Memory
• Director's Journal
• Critic's Review

Best for: AI Engineer, Machine Learning Engineer, AI Architect

Related on AIssential

• Why Multi-Agent Systems Need Memory Engineering — Multi-agent AI systems require robust memory engineering to ensure shared state consistency and prevent costly failures.
• Long-Running Agents — Long-running agents achieve sustained progress by externalizing state and decoupling components, overcoming single-session AI limitations.
• LAI #110: Fixing Context Rot and Rethinking How Agents Reason — Context rot, caused by noise burying essential details, is a primary reason for AI agent degradation in long tasks.
• Agent memory resolved? — A git-like memory framework significantly enhances coding agent performance and context persistence across sessions.
• Can LLM Agents Sustain Long-Horizon Organizational Dynamics? — Structured simulation memory is key for LLM agents to sustain coherent, long-horizon organizational dynamics and produce grounded artifacts.
• The Agent Stack - Part 5: Context, Retrieval, and Memory — Context is the runtime-assembled working set, not the model's inherent knowledge.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by InfoQ.