---
source: https://iinventedlight.com/library/notes/ao-coding-agentless-vs-agent-platforms.md
title: "Agentless pipelines vs full coding-agent platforms"
license: CC-BY-4.0
license_url: https://creativecommons.org/licenses/by/4.0/
source_license: "CC BY 4.0; restated in our own words"
attribution: "Summary by iinventedlight (AI-assisted). Sources: arXiv, [2407.01489] Agentless: Demystifying LLM-based Software Engineering Agents, CC BY 4.0; arXiv, [2407.16741] OpenHands: An Open Platform for AI Software Developers as Generalist Agents, CC BY 4.0."
author: "iinventedlight (synthesis of cited sources); text drafted with AI assistance"
date: 2026-11-04
status: published
checked_on: 2026-11-04
as_of: 2026-11-04
confidence: medium
untested: "No experiment was run; only the fetched arXiv abstract pages were read, so results are as reported by the authors."
category: agent-operations
topic_area: "agent-ops: coding agents"
tags: agents, coding-agents, swe-bench
sources:
  - id: S1
    title: "[2407.01489] Agentless: Demystifying LLM-based Software Engineering Agents"
    publisher: arXiv
    url: https://arxiv.org/abs/2407.01489
    accessed: 2026-11-04
    license: "CC BY 4.0"
  - id: S2
    title: "[2407.16741] OpenHands: An Open Platform for AI Software Developers as Generalist Agents"
    publisher: arXiv
    url: https://arxiv.org/abs/2407.16741
    accessed: 2026-11-04
    license: "CC BY 4.0"
---
# Agentless pipelines vs full coding-agent platforms

## Short answer
A simple fixed pipeline can match complex agents. Agentless uses three phases (localize, repair, validate the patch) without letting the model choose its next actions, and on SWE-bench Lite it reached 32% at about $0.70 per issue. Platforms like OpenHands instead give agents a sandboxed shell, code editing and a browser. Try the simple baseline before reaching for full autonomy. [S1][S2]

## Claims
- id: C1
  claim: Agentless uses localization, repair and patch validation.
  source: S1
  evidence: "a simplistic three-phase process of localization, repair, and patch validation"
- id: C2
  claim: Agentless reached the highest performance at low cost on SWE-bench Lite.
  source: S1
  evidence: "achieve both the highest performance (32.00%, 96 correct fixes) and low cost ($0.70)"
- id: C3
  claim: OpenHands supports safe interaction with sandboxed environments.
  source: S2
  evidence: "safe interaction with sandboxed environments for code execution"
- id: C4
  claim: OpenHands is released under the MIT license.
  source: S2
  evidence: "Released under the permissive MIT license"

## Sources
- [S1] [2407.01489] Agentless: Demystifying LLM-based Software Engineering Agents. arXiv. https://arxiv.org/abs/2407.01489 . License: CC BY 4.0. Accessed 2026-11-04.
- [S2] [2407.16741] OpenHands: An Open Platform for AI Software Developers as Generalist Agents. arXiv. https://arxiv.org/abs/2407.16741 . License: CC BY 4.0. Accessed 2026-11-04.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted).