awesome-harness-engineering: Find AI Agent Harness Resources

Overview

awesome-harness-engineering is a curated resource list about the systems surrounding AI agents: how they receive context, use tools, retain state, coordinate work, and verify results. It focuses on harness design rather than model selection, bringing together practical guidance, research, and implementation references.

The collection is useful as a map of a broad and evolving field, not as a software package or a prescriptive framework. Readers can browse by design problem, such as planning, permissions, observability, or evaluation, and follow links to the original resources.

Key Features

• Groups resources by harness foundations, design primitives, implementations, security, evaluation, templates, and operations.
• Covers practical concerns including agent loops, planning, context delivery, tool interfaces, memory, and orchestration.
• Includes references on MCP and skills, permissions, sandboxing, verification, and human oversight.
• Brings together tutorials, research papers, engineering articles, tools, and example harnesses.
• Supports exploratory learning across multiple approaches rather than prescribing a single agent framework.

Use Cases

• Agent developers can research a specific reliability issue, such as context limits, tool errors, or verification, before changing their harness.
• Engineering teams evaluating agent infrastructure can compare patterns and find reference implementations relevant to their requirements.
• Researchers and students can use the categorized links to discover papers and practical case studies on agent systems.
• Technical leads building production agents can survey security, evaluation, observability, and human-in-the-loop practices.

Project Facts

• Language: Python
• License: NOASSERTION
• Stars: 4.7k
• Forks: 618
• Topics: agent-harness, agent-memory, agent-orchestration, ai-agent-harness, ai-agents, awesome-list, context-engineering, harness-engineering, mcp
• Archived: false

Getting Started

Browse the repository README and follow the links under the topic that matches your question. This is a reference collection, so no installation is required.

Considerations

• The repository is a curated list, not an executable agent framework or a maintained implementation of the listed tools.
• Resource inclusion does not imply that a project is production-ready, compatible with your stack, or independently verified. Check each linked resource for its own status and requirements.
• The input reports the license as NOASSERTION; the README displays a CC0 badge, so confirm the repository's LICENSE file before reusing its contents.
• The collection is broad and actively changing, so readers may need to assess the relevance and freshness of individual links.