Repolode
Submit
ai-boost

ai-boost/awesome-harness-engineering

A curated awesome list of resources, patterns, and templates for designing the scaffolding around AI agents, covering memory, MCP, permissions, evals and orchestration.

View on GitHubWebsite
Stars
4,882
Forks
682
Open issues
262
Last commit
yesterday

What is it?

Awesome Harness Engineering is a curated list focused on the harness around an AI agent rather than the model itself. The harness covers context delivery, tool interfaces, planning artifacts, verification loops, memory systems, permissions and sandboxes, which determine whether an agent succeeds on real tasks.

The list is organized by design problem, including the agent loop, planning and task decomposition, context compaction, tool design, skills and MCP, permissions, memory and state, orchestration, verification, observability and human-in-the-loop. It also gathers reference implementations, tutorials, templates, evals and production case studies, with a foundations section of essays from OpenAI, Anthropic, Google, LangChain and academic papers.

It suits developers building or evaluating coding agents and other long-running agent systems who want to compare design approaches before choosing a framework.

Key features

  • 01Resources grouped by design primitives such as agent loop, planning, context compaction and tool design
  • 02Coverage of MCP, skills, permissions and memory management
  • 03Collection of evals, verification and observability references
  • 04Reference implementations, tutorials and reusable templates
  • 05Foundational essays and papers on agent architecture

Use cases

  • Designing the scaffolding for a coding agent or long-running agent workflow
  • Comparing harness design trade-offs before choosing an agent framework
  • Finding evaluation approaches and production case studies for agent reliability

Things to know

  • The README contains no installation or run instructions because this is a curated list of resources rather than runnable software.
  • The license is listed as NOASSERTION, so reuse terms are unclear from the metadata.

Topics

#ai-agents#agent-harness#llm#mcp#evals#agent-memory#orchestration#awesome-list#agent-orchestration#ai-agent-harness

FAQ

What is awesome-harness-engineering?
A curated awesome list of resources, patterns, and templates for designing the scaffolding around AI agents, covering memory, MCP, permissions, evals and orchestration. Awesome Harness Engineering is a curated list focused on the harness around an AI agent rather than the model itself. The harness covers context delivery, tool interfaces, planning artifacts, verification loops, memory systems, permissions and sandboxes, which determine whether an agent succeeds on real tasks.
Is awesome-harness-engineering free to use? What license does it have?
awesome-harness-engineering is open source under the Other license. Check the license terms before using it in commercial products.
Is awesome-harness-engineering still maintained?
Actively maintained. The last commit to awesome-harness-engineering was on Oct 19, 2026.
What can I use awesome-harness-engineering for?
Designing the scaffolding for a coding agent or long-running agent workflow Comparing harness design trade-offs before choosing an agent framework Finding evaluation approaches and production case studies for agent reliability
Where can I get awesome-harness-engineering?
The source code and installation instructions are on GitHub: https://github.com/ai-boost/awesome-harness-engineering

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects