danielrosehill/AI-Harnesses — Summary

AI-Harnesses by Daniel Rosehill is a point-in-time snapshot (April 2026) of projects self-describing as AI agent harnesses or related to "harness engineering." It is a curated list with a README-only structure (no subdirectories, no code) containing two tables: "Agent Harnesses" (21 runtime projects) and "Resource Lists & Guides" (5 catalog/guide projects). The README's Note On Terminology section is notable: it documents the etymology and emergence of "agent harness" as a term in late 2025/early 2026, the relationship to "harness engineering" as a practice, and the ambiguity between harness/framework/platform. Star counts are as of April 2026 (static, will not be updated). With 3 stars, this is primarily a research artifact for the author rather than a community resource — but its taxonomy (harness runtime vs. resource list) and etymology section are useful for Phase B analysis.

differs_from_seeds: No direct seed analog — static snapshot/taxonomy document. Unique in documenting the etymology of "harness engineering" as a term emerging in late 2025/early 2026. Among batch-27 peers, it is the most minimal documentation entry (single README, no content depth) but also the only one that explicitly addresses terminology ambiguity and the harness/framework/platform distinction.

danielrosehill/AI-Harnesses — Overview

Origin

Created by Daniel Rosehill. No license. 3 stars, 0 forks. All Markdown. Default branch: master. Last commit 2026-04-04. Self-described as "Point-in-time snapshot of projects describing themselves as AI agent harnesses (April 2026)."

Terminology Section (key value of this repo)

Verbatim from README:

The term "agent harness" emerged in late 2025 / early 2026 as a way to describe the orchestration layer that runs an AI agent. A harness manages the tool-dispatch loop, permissions, context, and lifecycle of an agent — as distinct from a framework (which provides higher-level abstractions for building agents) or a backend (which provides the inference and model serving).

The term gained traction partly through the adjacent concept of "harness engineering" — the practice of configuring and optimizing these orchestration layers (writing CLAUDE.md files, designing tool schemas, structuring agent prompts, etc.).

This list captures projects that self-identify with the "harness" label as of April 2026. The terminology is still fluid and not universally agreed upon — some projects listed here blur the line between harness, framework, and agent platform.

Two Tables

Agent Harnesses (21 projects): Runtime orchestration projects — deer-flow, oh-my-openagent, langchain-ai/deepagents, aden-hive/hive, xerrors/Yuxi, mindfold-ai/Trellis, HKUDS/OpenHarness, can1357/oh-my-pi, autoagent, alphaclaw, cc-mini, dexto, claude-code-harness, PantheonOS, open-harness, Autono, AutoHarness, srchd, OpenHarness-For-Codex, mastersof-ai/harness

Resource Lists & Guides (5 projects): walkinglabs/awesome-harness-engineering, AutoJunjie/awesome-agent-harness, Picrew/awesome-agent-harness, alchemiststudiosDOTai/harness-engineering, jrenaldi79/harness-engineering

Key Insight

The Resource Lists section is a meta-catalog of catalogs — this repo is a meta-meta-catalog listing other curated lists including awesome-harness-engineering-walkinglabs from this same batch.

danielrosehill/AI-Harnesses — Architecture

Distribution

• Type: methodology-doc (static taxonomy snapshot)
• Format: Single README.md
• License: none
• Branch: master (not main)

Directory Structure

README.md              # The entire content

(Only one file in the entire repository)

Required Runtime

None.

About the Snapshot

Static snapshot as of April 2026. Author states: "Star counts as of April 2026. This is a static snapshot and will not be updated."

This means the repository is intentionally frozen — a research artifact, not a maintained list.

danielrosehill/AI-Harnesses — Components

This is a single-file taxonomy snapshot.

Agent Harnesses Table (21 projects)

Resource Lists Table (5 projects)

Note: walkinglabs/awesome-harness-engineering now has 2,734 stars (as of May 2026 analysis) — grew 2.2x since this April 2026 snapshot.

danielrosehill/AI-Harnesses — Uniqueness & Positioning

differs_from_seeds

Tier C by complexity — single-file static taxonomy snapshot. No direct seed analog. Unique value: the terminology section documenting the emergence of "agent harness" as a term in late 2025/early 2026, the distinction from "framework" and "platform," and the acknowledged fluidity of the terminology. The April 2026 star counts provide a temporal baseline: walkinglabs/awesome-harness-engineering had 1,264 stars at snapshot time (now 2,734 — grew 2.2x in ~6 weeks). This makes it useful as a historical marker, even if the content is minimal.

Positioning

• Research artifact / personal taxonomy note
• Etymology documentation for the "harness" term
• Cross-reference point: lists walkinglabs/awesome-harness-engineering (now 2,734 stars) at 1,264 stars (April 2026 baseline)

Why Not Tier C

Despite minimal content, the repo provides two things that matter for Phase B analysis:

1. The most precise single-paragraph disambiguation of harness vs. framework vs. platform in the corpus
2. A temporal star-count baseline that reveals the rapid growth of this category (the field went from niche to mainstream between April and May 2026)