Awesome AI AgentsAI Agent Development

ruvnet/metaharness

⭐ 652 TypeScript repository created 2026-06-13

MetaHarness describes itself as a factory for agent frameworks rather than another agent framework. Given a GitHub URL or a blank slate, it mints a repository-aware harness: a CLI published under the user chosen name, a repository-aware coding agent, a local MCP server, a memory namespace scoped to the project, skills derived from the actual file layout, a governance policy, release verification gates and witness-signed provenance. The output is an npm-publishable archive carrying the author own branding. Generation runs from the command line through npx or from a browser studio, and the project states that nothing leaves the machine during scaffolding. A generated harness can target one or several hosts, including Claude Code, OpenAI Codex, pi.dev, Hermes, OpenClaw, RVM and Prime Agent, with host-specific adapters translating tools into whatever the host understands; the Prime Agent adapter, for example, emits Python-backed project-scoped skills because that host has no MCP support, and ships an explicit sandbox-required note when a deny-list cannot be enforced. A scoring command reads a repository without executing it and reports how well a harness would fit, how likely the build is to succeed, how safe the tools are and the approximate cost per run. Companion packages extend the generated harness: a router that picks the cheapest model meeting a quality bar from the user own results, a self-improvement mode that mutates harness configuration in a sandbox and keeps only changes that measurably improve benchmark scores while the model stays frozen, a governed autonomous variation runtime with budgets, quarantine, rollback and signed replay receipts, and an experimental ARC-AGI-3 controller exposed as a remote MCP server. The README marks several performance claims as deferred behind preregistered evaluation gates rather than asserting them.

https://github.com/ruvnet/metaharness

agent-harnessagent-scaffoldingagentic-aiautonomous-agentsclaude-codecodexmcpmcp-servermulti-agentsandboxingcode-generationtypescript

Also in AI Agent Development

kepano/obsidian-skills

Obsidian Skills provides a collection of agent skills enabling AI agents to interact with Obsidian vaults, facilitating content creation and editing of Markdown, Bases, and JSON Canvas files, as we...

Panniantong/Agent-Reach

Agent-Reach allows AI agents to access web content from major social media and video platforms through a CLI, without API fees, enhancing their perceptual capabilities.

JCodesMore/ai-website-cloner-template

This project provides a reusable Next.js template for reverse-engineering and cloning any website into a modern codebase using various AI coding agents.

huggingface/smolagents

Smolagents is a minimalistic Python library by Hugging Face for creating intelligent agents that think and act by generating and executing Python code, supporting multiple LLMs, modalities, and tool integrations.

alibaba/page-agent

Page Agent is a JavaScript in-page GUI agent enabling natural language control of web interfaces without browser extensions or headless browsers, supporting various LLMs.

TencentCloud/TencentDB-Agent-Memory

TencentDB Agent Memory provides a local, 4-tier progressive pipeline for AI agent long-term memory, enhancing performance and reducing token usage without external API dependencies.

titanwings/distilly

A meta-skill that distills messages, documents, interviews and public sources about a person into a source-grounded Person Profile, packaged as an Agent Skill for coding agents and chat bots.

coze-dev/coze-studio

Coze Studio is an all-in-one AI agent development platform that simplifies the creation, debugging, and deployment of AI agents through visual tools and supports no-code and low-code development approaches.