Foundational Papers

Key papers that shaped the field of Harness Engineering and AI agent runtimes.

Agent Architectures

  • ReAct: Synergizing Reasoning and Acting in Language Models (2022) Yao et al. | Paper The reasoning + acting loop that underlies most modern agent harnesses.

  • Toolformer: Language Models Can Teach Themselves to Use Tools (2023) Schick et al. | Paper Foundation for tool-use in agent harnesses.

  • Generative Agents: Interactive Simulacra of Human Behavior (2023) Park et al. | Paper Memory architecture for persistent agent identities.

  • The Landscape of Emerging AI Agent Architectures for Reasoning, Planning, and Tool Calling (2024) Masterman et al. | Paper Comprehensive survey of agent architecture patterns.

Memory & Context

  • MemGPT: Towards LLMs as Operating Systems (2023) Packer et al. | Paper Virtual context management inspired by OS memory hierarchy.

  • Reflexion: Language Agents with Verbal Reinforcement Learning (2023) Shinn et al. | Paper Self-reflection as a memory and learning mechanism.

Multi-Agent Systems

  • AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation (2023) Wu et al. | Paper Multi-agent conversation framework design patterns.

  • AgentVerse: Facilitating Multi-Agent Collaboration (2023) Chen et al. | Paper Coordination patterns for multi-agent systems.

Harness Engineering (Emerging)

  • Harrison Chase: "Your Harness, Your Memory" (2026) Blog Post | Thread The essay that put "Harness Engineering" into mainstream discourse.

  • steipete: OpenClaw Strict Mode & Harness Pluggability (2026) Thread Runtime-level constraints vs prompt-level constraints.

This list is actively maintained. Submit papers via PR or open an issue with the paper label.