The cross-harness, self-improving memory layer for AI agents. Join our community: https://discord.com/invite/zdNChS2fBu
-
Updated
Nov 20, 2026 - Go
Agent harnesses are the runtime scaffolding around AI agents. They usually combine context delivery, tool interfaces, planning state, memory, sandboxes, permissions, evaluation, and observability so agents can complete longer tasks reliably. Agent harnesses are especially common in coding agents, research agents, and multi-agent workflows where repeatability, safety, and traceability matter.
The cross-harness, self-improving memory layer for AI agents. Join our community: https://discord.com/invite/zdNChS2fBu
HelixHarness 提供完整的 AI Coding 工程化基础能力,还支持不同企业和团队结合自身的业务知识、工程规范和研发流程,持续构建和演进属于自己的专属 Harness
Recursive Experiential–Working Memory Evolution for Long-Horizon Agent Harnesses
Open-source Python agent harness for auditable, cost-controlled LLM workflows with YAML playbooks and MCP server support.
agent-harness-engineering-pattern
Mira 是一款本地优先的个人 AI 桌面工作台,让智能体通过文件、终端、浏览器和电脑控制工具完成任务,支持多模型接入、智能体协作、工作流自动化与 MCP 扩展。
A curated list of open source and source available agent harnesses
Progressive-disclosure skill management across Codex, Claude Code, Cursor, Gemini, Hermes, and other agent harnesses.
expert skill package for AI coding assistants working with the t1yOS Serverless platform.
Evidence-first Claude Code internals research: Agent Loop, Harness design, failure recovery, and reproducible experiments.
Insi measures whether a coding agent gets better at a repository over time, and is built to be added to any repository as a dependency.
A local workbench for recording, replaying, comparing, and inspecting tool-using agent runs without sending trace data to a hosted service.
Open-source framework for reproducible A/B evaluation of agent skills and instructions
DeepSeek Harness's desktop client.
🧩 Personal agent skills shared across AI coding tools and harnesses.
Local-first macOS menubar AI cockpit & dual-role MCP multiplexer. Connects local OpenAI-compatible inference runtimes and agent harnesses to your Obsidian vault knowledge graph for sovereign memory. Real-time text telemetry, connection graph visualization, zero-config tool routing, and local RAG search.
Self-updating intelligence dashboard tracking the AI model and tooling landscape — frontier models, agent harnesses, self-hosting, strategy.
Interlinked markdown wiki on agent harnesses, orchestration, formal methods, and adjacent research.
Codex skill for designing and scaffolding durable agent harness projects, with a Ralph Loop preset and upgrade-friendly doctrine.
Bounded, receipt-producing task handoffs between independently authenticated local agent harnesses.