#agent-harness

HarnessRouter Community Edition: the self-hosted, Apache-2.0 edition of the unified interface for agent harnesses. Run Codex, Claude Code, Hermes, PI, DSH, and more through one API, with sessions, streaming, files, cancellation, and failure handling. Implements the Unified Harness Protocol (UHP), an open standard. Your keys, your infrastructure.

An agent harness that compiles a model into one provably-correct, self-retargeting CUDA megakernel and self-tunes it past cuBLAS at batch-1 LLM decode, paper: https://arxiv.org/abs/2606.09682

Finish-First Autonomous Agent Loop for Claude Opus 4.7 – 2026 Edition

CubeLoop — a Pythonic, async-native agent harness core designed for high performance, readability, and production-grade persistence.

Fable-style spec + evidence gate for Claude Code + Codex. Makes Opus/Codex work under Fable-like discipline: blocks every edit until a deterministic spec passes, and there is no "done" without live acceptance evidence. Spec-first, verification-gated, forbidden-paths enforced.

Skill repository to enforce Engineering disclipline in coding assistants

AI-assisted XAUUSD daily research workflow with Research Skill, Agent Harness, LangGraph, Tool Calling, and free public data sources.

The control plane for your coding agents, however you run them. One checkout of rules, skills, subagents, hooks and stances, projected into Claude Code and Codex with your own primitives: hook-enforced guardrails, usage telemetry, cost postures and model tiering, under any rules library or orchestration layer.

Local agency/runtime harness for ChatGPT Web: repositories, tasks, commands, validation, Git, skills/memory, observability, and opt-in computer control.

Aielia — an open-source everyday chat assistant that runs a full 11-layer agent harness (world model, control-state gate, verification, recovery, learning) on every turn. Plus a visual canvas that compiles to LangGraph / CrewAI / Mastra / MS Agent Framework.

An interactive roadmap for harness engineering — building the agent loop, tool layers, context engineering, memory, retrieval, eval harnesses, and the rest of the scaffolding around modern LLMs.

Measure a repository as a place for a coding agent to work, and lay the foundation where it has none. A Claude Code plugin: hooks, guards and gates proven to turn red, layered docs, a repo graph, and a five-dimension assessment you can take again later.

Open-source control plane for coding agents you own — run Codex, Claude Code, Gemini, Cursor and ACP agents side by side with shared history, permissions, handoffs and audit logs.

OS-level safeguards layer for AI agent loops — runaway-kill, timeouts & resource limits so coding agents (Claude Code, Cursor, Codex, Antigravity) in YOLO mode can't burn tokens or freeze your Mac

An opinionated terminal agent with durable sessions, background jobs, and first-class unattended runs. Written in Go.

IDE skills for scaffolding AI agent harnesses: ReAct, plan-and-execute, self-evolving, reflexion, multi-agent, code agent, RAG memory

Local execution harness for OpenAI Codex with durable ledgers, verification gates, memory, and replay-gated skills.

IssueOps: issue-driven agent workflows for Codex, Claude Code, and Omo

Native DeepSeek Harness (DSH) plugin integrating TypeSafe Jev as a System One decision layer for agent selection, supervision, corrections, and approvals.

A local-first coding agent: runs the heavy executor on your local model and escalates to the cloud only when stuck. Out-resolves single-shot Opus/Sonnet by planning, running the project's real tests, and retrying — at ~half the cost of an all-cloud chain. Graded by SWE-bench, not an LLM.

面向多租户的 Agent Harness / Runtime 参考实现 —— 凭证代管、权限控制、多租户隔离、可回归 eval

Build production-ready travel planning agents with a trained 4B model, tool-calling harness, and evidence-backed itineraries.

Claude Code, Codex, and multi-agent configuration system: principles, hooks, skills, and workflow patterns for AI-assisted development

A provider-agnostic scaffolding kit for running structured multi-agent workflows in your codebase.

A Laravel agent harness with swappable loops. Pick how your AI agent thinks, switch with one line, compare which works best.

A lightweight local Coding Agent harness inspired by learn-claude-code s20, featuring a minimal LLM tool-use loop, workspace-safe file access, command execution, and pytest-verified Phase 1 tools.

A generator-critic harness for Claude Code: Planner / Executor / a parallel panel of tool-backed jurors / a routing Judge, with candidate isolation and CI as the final gate.

DeepSeek-first local Windows agent for files, research, automation, verified computer use, and auditable office workflows.

A lightweight, explicit, Git-native coding agent continuity and governance harness.

Agent Harness: whole-codebase review harness — parallel correctness, security, and reuse channels.

Agent Harness: field-level module audit harness — Excel workbook + proven valid/invalid tests.

Agent Harness: visual demo-readiness harness — is this screen understandable in ten seconds?

Agent Harness: manual UI QA harness — SHIP/DON'T-SHIP verdict through the real running app.

Agent Harness: cross-cutting impact harness — 7-check sweep for path divergence, cache drift, staleness.

Agent Harness: merge-review harness for Cursor and Claude Code — source-verified GO/NO-GO before merge.

Model Orchestrator routes coding-agent work across Codex, Claude Code, OpenCode, OpenClaw, and Hermes with measured effort lanes, worker profiles, and a transparent Routing Studio.

One shared working contract for every AI coding assistant you run: Claude Code, Codex, Cursor, opencode, dsh and oh-my-pi (omp).

Canary — a minimal, modular, self-improving agent harness with atomic releases, shared memory and an OpenAI-compatible API

A lightweight, native Windows AI agent — WinUI 3 + .NET 9, works with any OpenAI-compatible endpoint.

Halo Harness: an intuitive, Claude Code-compatible agent harness that drives DeepSeek, GLM, Kimi, Qwen and Claude over Databricks, OpenRouter, the Anthropic API and your Claude Code subscription, reusing your existing Claude Code configuration.