Repolode
Submit
RyanAlberts

RyanAlberts/best-of-Agent-Harnesses

A curated, ranked directory of 187 AI agent harnesses, with templates, playbooks, skills, and an MCP server for choosing one.

View on GitHubWebsite
Stars
1,464
Forks
99
Open issues
7
Last commit
3 days ago

What is it?

This repository is a curated, regularly rescored list of AI agent harnesses: the runtime layer that lets a language model call tools in a loop, decide what needs approval, and manage context and crash recovery. Projects are grouped into categories such as coding agent products, personal agent runtimes, frameworks, multi-agent orchestration, memory, tool discovery, and evaluation benchmarks, each with star counts and a rough complexity tier.

Beyond the catalog, it offers comparison pages for head-to-head decisions (for example terminal coding agents, memory layers, or browser agents), copy-ready templates such as a shared AGENTS.md and safe Claude Code settings, and step-by-step playbooks for building a minimal harness. It also includes installable skills that inspect a user's own agent setup for guardrail gaps, token waste, and unverified test claims.

The list is published as harnesses.json, llms.txt, and an MCP server, so coding agents can query it to recommend a harness for a given model and task.

Key features

  • 01Ranks 187 agent harnesses by category, complexity tier, and stars
  • 02Provides head-to-head comparison pages for terminal coding agents, memory layers, and browser agents
  • 03Ships copy-ready templates such as a shared AGENTS.md and safe Claude Code permission settings
  • 04Offers installable skills that audit guardrails, token usage, and test claims in your own agent setup
  • 05Publishes machine-readable data and an MCP server so agents can query recommendations

Use cases

  • Choosing a coding agent or agent framework before starting a new project
  • Comparing memory, browser, or multi-agent tools side by side for a specific task
  • Setting up a shared instruction file and safe permission settings across several coding agents
  • Letting an AI assistant recommend a harness through the MCP server

Things to know

  • Content is licensed CC-BY-SA-4.0, so reuse must credit the author and share adaptations under the same license.

In collections

Open-source AI agent frameworks & appsAI coding assistants & agentsMCP servers

Topics

#ai-agents#llm#awesome-list#coding-agents#mcp#agent-framework#developer-tools#agent-harness#agentic-ai#agents

FAQ

What is best-of-Agent-Harnesses?
A curated, ranked directory of 187 AI agent harnesses, with templates, playbooks, skills, and an MCP server for choosing one. This repository is a curated, regularly rescored list of AI agent harnesses: the runtime layer that lets a language model call tools in a loop, decide what needs approval, and manage context and crash recovery. Projects are grouped into categories such as coding agent products, personal agent runtimes, frameworks, multi-agent orchestration, memory, tool discovery, and evaluation benchmarks, each with star counts and a rough complexity tier.
Is best-of-Agent-Harnesses free to use? What license does it have?
best-of-Agent-Harnesses is open source under the CC-BY-SA-4.0 license. Check the license terms before using it in commercial products.
Is best-of-Agent-Harnesses still maintained?
Actively maintained. The last commit to best-of-Agent-Harnesses was on Nov 3, 2026.
What can I use best-of-Agent-Harnesses for?
Choosing a coding agent or agent framework before starting a new project Comparing memory, browser, or multi-agent tools side by side for a specific task Setting up a shared instruction file and safe permission settings across several coding agents Letting an AI assistant recommend a harness through the MCP server
Where can I get best-of-Agent-Harnesses?
The source code and installation instructions are on GitHub: https://github.com/RyanAlberts/best-of-Agent-Harnesses

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

Related projects