Topics / #harness-engineering

Harness-engineering projects on GitHub

Building and evaluating agent harnesses and scaffolding.

As of , the most-starred Harness-engineering projects on GitHub are ragflow (92.4K stars), learn-claude-code (78.7K stars) and QwenPaw (35.8K stars). GitHub.Locker tracks 38 Harness-engineering repositories with 428.4K stars combined. The fastest-growing on October 16 was learn-harness-engineering, with +115 stars in a day.

38Projects
428.4KStars in total
+566Stars in a day
Browse in the catalog →github.com/topics/harness-engineering ↗

Hot right now

Most stars gained in a day · Oct 16

+115

learn-harness-engineering

walkinglabs/learn-harness-engineering · ★ 20K

A free, project-based course on harness engineering: building the environment, state management, verification and control that make AI coding agents complete real repo work reliably. It has lectures, hands-on projects and a template library, translated into 15 languages.

+65

learn-claude-code

shareAI-lab/learn-claude-code · ★ 78.7K

It is an educational repository that teaches how to build an 'agent harness' (the operational environment around an LLM) from scratch, using Claude Code as the reference implementation, across 17 progressive Python lessons.

+55

ragflow

infiniflow/ragflow · ★ 92.4K

RAGFlow is an open-source Retrieval-Augmented Generation (RAG) engine that combines RAG with Agent capabilities to build a context layer for LLMs, turning complex enterprise data into production-ready AI systems.

+55

harnessrouter

HarnessRouter/harnessrouter · ★ 3.1K

HarnessRouter is a self-hosted gateway that turns existing agent harnesses like Codex, Claude Code, Hermes, PI, and DSH into a single unified API, so a product can run tasks, stream progress, and switch harnesses without handling harness-specific engineering.

The all-time top

38 projects · by stars

  1. #1 · infiniflow/ragflow

    ragflow

    RAGFlow is an open-source Retrieval-Augmented Generation (RAG) engine that combines RAG with Agent capabilities to build a context layer for LLMs, turning complex enterprise data into production-ready AI systems.

    92.4K
  2. #2 · shareAI-lab/learn-claude-code

    learn-claude-code

    It is an educational repository that teaches how to build an 'agent harness' (the operational environment around an LLM) from scratch, using Claude Code as the reference implementation, across 17 progressive Python lessons.

    78.7K
  3. #3 · agentscope-ai/QwenPaw

    QwenPaw

    QwenPaw is a personal AI assistant that you deploy yourself (locally or in the cloud), which remembers context across sessions, runs tasks/automations, and reaches you through chat apps like Discord, Telegram, and WeChat. It is extensible with Skills, Plugins, and MCP tools.

    35.8K
  1. 4

    deepagents

    30.3Klangchain-ai/deepagents · Python

    Deep Agents is an open source, batteries-included agent harness from LangChain: an opinionated agent that runs out of the box with sub-agents, a filesystem, context management, and skills bundled in — and every piece can be extended, overridden, or replaced without forking.

  2. 5

    symphony

    27.7Kopenai/symphony · Elixir

    Symphony is an experimental system from OpenAI that turns project tasks into isolated, autonomous implementation runs, so teams manage the work itself instead of babysitting individual coding agents.

  3. 6

    pydantic-ai

    20.7Kpydantic/pydantic-ai · Python

    Pydantic AI is the Python AI SDK for building agents: a typed, extensible agent loop where every model is a string swap away. The same agent runs behind a web frontend, in the terminal, on a voice call, or as a plain object you call run() on.

  4. 7

    learn-harness-engineering

    20Kwalkinglabs/learn-harness-engineering · TypeScript

    A free, project-based course on harness engineering: building the environment, state management, verification and control that make AI coding agents complete real repo work reliably. It has lectures, hands-on projects and a template library, translated into 15 languages.

  5. 8

    DeepCode

    16.7KHKUDS/DeepCode · Python

    An open agentic coding platform from HKUDS: a multi-agent harness that turns natural language and even research papers into production-ready code, working directly in your repository.

  6. 9

    hive

    11.1Kaden-hive/hive · Python

    OpenHive is an open-source multi-agent harness for production AI — a runtime that adds state management, failure recovery, observability, cost control, and human oversight so colonies of agents reliably run real business processes instead of demos.

  7. 10

    harness

    9.2Krevfactory/harness

    Harness is a Claude Code plugin that turns a plain-English description of your project into a custom team of specialized AI agents and the skills they use, chosen from six pre-defined team-architecture patterns.

  8. 11

    munder-difflin

    8.8KHarnessMD/munder-difflin · TypeScript

    Munder Difflin is a free, open-source desktop app that turns the coding-agent CLIs you already run (Claude Code, Codex, Gemini CLI, Grok, Qwen, OpenCode and more) into a coordinated office of autonomous agents, visualized as avatars working on a 2D floor and managed by your own AI clone, Michael.

  9. 12

    defending-code-reference-harness

    7.6Kanthropics/defending-code-reference-harness · Python

    A reference implementation from Anthropic for finding and fixing vulnerabilities with Claude: a set of Claude Code skills (quickstart, threat-model, vuln-scan, triage, patch, customize) plus an autonomous scanning harness that runs a recon-find-verify-report-patch loop, built from learnings partnering with security teams.

  10. 13

    claw-code-parity

    6.6Kultraworkers/claw-code-parity · Rust

    A parity port of Project Claw Code — a rebuild of the Claude Code coding harness whose exposed snapshot the author chose not to track. The stated goal: prove an open coding harness can be built autonomously, in public, at high velocity.

  11. 14

    trueforge

    6.3Ktruefoundry/trueforge · TypeScript

    TrueForge is an open-source agent harness - the runtime layer that runs the full agent execution loop (model calls, MCP tools, skills, sandboxing, approvals, context management, session state) and exposes it via a chat UI, an HTTP API with a TypeScript SDK, and an embeddable UI SDK.

  12. 15

    ouroboros

    6.2KQ00/ouroboros · Python

    Ouroboros is an 'Agent OS' for AI coding — a local-first runtime layer that turns non-deterministic coding-agent work into a replayable, observable, policy-bound workflow (interview, crystallize, execute, evaluate, evolve) instead of ad-hoc prompting.

  13. 16

    harness-engineering

    6.1Kdeusyu/harness-engineering · JavaScript

    It's a Chinese-language deep-learning archive about 'Harness Engineering' — the OpenAI-coined engineering paradigm where engineers stop writing code directly and instead design constraints, feedback loops, and environments so AI agents can work reliably. The repo tracks the author's journey from reading the original essay to forming opinions, running experiments, and producing translated research.

  14. 17

    aidlc-workflows

    5.4Kawslabs/aidlc-workflows · TypeScript

    AI-DLC turns AI coding assistants into structured, verifiable software-delivery workflows, running a single harness-neutral core across Claude Code, Kiro, Codex CLI, Cursor, opencode, and GitHub Copilot.

  15. 18

    hermes-agent-orange-book

    5Kalchaincyf/hermes-agent-orange-book · Python

    This is a free ebook (the Orange Book series) that explains Hermes Agent, an open-source AI Agent framework from Nous Research, covering its self-improvement loop, memory system, tools, and multi-agent orchestration.

  16. 19

    awesome-harness-engineering

    4.9Kai-boost/awesome-harness-engineering · Python

    A curated list of resources on 'harness engineering' — the discipline of designing the scaffolding (context delivery, tool interfaces, planning artifacts, verification loops, memory, sandboxes) around an AI agent that determines whether it succeeds on real tasks.

  17. 20

    autoagent

    4.6Kthirdlayerinc/autoagent · Python

    AutoAgent lets an AI agent build and iterate on an agent harness (agent.py) autonomously overnight: it tweaks the system prompt, tools, config, and orchestration, scores each change against a benchmark, and keeps only the improvements.

  18. 21

    awesome-harness-engineering

    4.4Kwalkinglabs/awesome-harness-engineering

    A curated list of articles, playbooks, benchmarks, specifications, and open-source projects about harness engineering — shaping the environment around AI agents so they can work reliably.

  19. 22

    claude-code-book

    4.3Klintsinghua/claude-code-book · Python

    'Yu Yu: Decoding Agent Harness' (御舆:解码 Agent Harness) is a roughly 420,000-character Chinese technical book that reverse-engineers how Claude Code's Agent Harness works internally - the runtime driving the conversation loop, tool calls, permission checks, memory, context management and multi-agent orchestration.

  20. 23

    modlens

    4.3Kliustack/modlens · TypeScript

    ModLens is a vision plugin for DeepSeek Harness (dsh) that gives text-only coding models like DeepSeek and GLM the ability to see. Paste an image into the chat and it returns structured JSON evidence from OCR, layout, and semantic analysis.

  21. 24

    goclaw

    3.7Knextlevelbuilder/goclaw · Go

    GoClaw is a multi-tenant AI agent gateway/platform written in Go that connects 20+ LLM providers to 7 messaging channels, letting you deploy and orchestrate teams of AI agents at scale with production-grade security.

  22. 25

    harnessrouter

    3.1KHarnessRouter/harnessrouter · Python

    HarnessRouter is a self-hosted gateway that turns existing agent harnesses like Codex, Claude Code, Hermes, PI, and DSH into a single unified API, so a product can run tasks, stream progress, and switch harnesses without handling harness-specific engineering.

  23. 26

    GLM-5.3-Flash-J-Space-Capability-Realization-Report

    1KTiger3807861189/GLM-5.3-Flash-J-Space-Capability-Realization-Report

    J-Space Cognition Suite is a model-agnostic inference-time control system for deep reasoning, long-horizon work, tool use, verification, and recovery, packaged as a Skill that loads into AI agent hosts.

  24. 27

    ai-engineer-notebooks

    700calmrocks/ai-engineer-notebooks · Jupyter Notebook

    A collection of framework-free, hands-on Jupyter/Colab notebooks teaching the applied-LLM engineer skill set — model APIs, RAG, evals, agents, fine-tuning, security, and LLMOps — built for engineers moving into AI Engineer / Forward Deployed Engineer (FDE) roles.

  25. 28

    JIT

    556bingreeky/JIT · Python

    JIT-Agent is a compact meta-agent that writes a task-specific agent harness on the fly instead of relying on one precompiled general-purpose scaffold, wrapping any off-the-shelf agentic LLM in what the authors call 'Model-as-a-Harness'.

  26. 29

    Awesome-Graph-Engineering

    382DEEP-JLU/Awesome-Graph-Engineering

    A curated collection of research papers, benchmarks, and open-source projects accompanying a survey on Graph Engineering for LLM agents, covering the progression from Model Intelligence to Individual Intelligence to System Intelligence.

  27. 30

    eval

    310frontier-harness-eval/eval · JavaScript

    FrontierHarness Eval is a public benchmark that ran the same Kimi K3 model through nine coding-agent harnesses (12 configurations) on the same 30 software-engineering tasks, to see how pass rate, cost, cache behavior and speed change when only the harness changes.

  28. 31

    HarnessEval-W

    307MirroS-Lab/HarnessEval-W · Python

    HarnessEval-W is an agentified evaluation pipeline for benchmarking world/video-generation models: instead of a brute-force scalar score, it decomposes each evaluation case into sub-questions answered by specialized sub-agents, then aggregates the evidence into a transparent, auditable verdict.

  29. 32

    dsh-tianshu-tui

    287huiliyi37/dsh-tianshu-tui · TypeScript

    dsh-tianshu-tui is an interactive terminal UI (TUI) plugin for the official DeepSeek Harness CLI (@deepseek-ai/dsh). It uses a self-built minimal ANSI rendering engine to show streaming Markdown and tool cards, 16+ themes, slash commands, memory, and LSP diagnostics inside your terminal.

  30. 33

    zhijian-ai-bluebook-workbuddy-harness

    240zjp1997720/zhijian-ai-bluebook-workbuddy-harness

    This is a book (Zhijian AI Bluebook ZJBB-003) that reverse-engineers WorkBuddy, an AI agent app, tracing its UI behavior back to the local files it writes and the requests it sends to the model, covering prompts, memory, plugins, experts, Skills and security boundaries.

  31. 34

    oar

    198botiverse/oar · TypeScript

    OAR (Open Agent Runtime) is a provider-independent programming interface for coding-agent harnesses like Claude Code, Codex and Cursor, so you build one integration instead of writing separate harness logic for each agent.

  32. 35

    SystemOneHarness

    194HarnessRouter/SystemOneHarness · Python

    An open-source Python harness that turns a System One decision model into an agent loop: it observes an environment, compiles a finite action space into typed questions, gates each decision by confidence, executes the action and records a complete trace.

  33. 36

    Benzi

    163oooscoos/Benzi · HTML

    Benzi is a compiler-backed coding agent: it parses your whole codebase with tree-sitter into a queryable index of symbols, calls, and data/control flow, then uses that map to explain, edit, and verify code changes, including a runtime tracer that settles what static analysis alone can't.

  34. 37

    awesome-ai-harness

    153weiwei966/awesome-ai-harness

    A curated reading list on 'harness engineering' — the runtime scaffolding (context management, tool design, agent loops, memory, sandboxing, evaluation) that turns a raw language model into a working agent, in English and Chinese.

  35. 38

    Proteus

    108proteus-evolve/Proteus · Python

    Proteus lets you plug any agent harness into any model, have it rewrite its own harness (prompts, memory, skills, tools, control loop) across many context-fresh episodes, and measures how that harness actually changes over time.

Often tagged together