Article:Meng Li
DeepSeek Harness vs. Codex Harness: Competing Engineering Philosophies for AI Agents
In August 2026, OpenAI and DeepSeek both introduced execution harnesses for AI agents. Meng Li examines how their technical philosophies diverge.
OpenAI positions Codex as a platform powered by an open agent harness
ArticleOpenAIAnnouncement
OpenAI highlights its open-source Codex harness as a platform allowing developers to embed AI agents directly into specialised workflow tools rather than forcing tasks into generic chat interfaces.
Key points
- The open-source harness includes the Codex CLI, official Codex SDK, and Codex app-server, while model access remains separate.
- The harness layer manages conversation state, tool calls, streaming execution, sandboxing, and approval policies.
- OpenAI notes that retained reasoning and context compaction increased GPT-5.6 Sol's ARC-AGI-3 benchmark score from 13.3% to 38.3% while cutting output tokens sixfold.
- Three integration layers are offered: codex exec for noninteractive scripts/CI, the Codex SDK for programmatic tasks, and app-server for interactive in-product agents.
- Early adopters include GitHub/JetBrains, Cisco Cloud Control, and tax-preparation workflows at Thrive Holdings and Crete.
Why it matters
Rather than relying on generic chat boxes, developers can retain their domain-specific interfaces, queues, and approval gates while embedding agentic capabilities connected via tools like MCP.
Analysis
By open-sourcing the agent harness and app-server protocol, OpenAI provides an orchestration layer between applications and models. This architecture lets teams keep existing business UI and compliance controls while outsourcing the complex loop of turn management, streaming, and tool execution to standard components.
Mentioned links
Found in
Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.
Related entries
Similar in content, found by the AI search.
Article:learn.chatgpt.com
ChatGPT and Codex: Unified Plugin Ecosystem with MCP and Connectors
OpenAI outlines the universal plugin architecture for ChatGPT and Codex, enabling reusable workflows with skills, MCP servers, connectors, and lifecycle hooks across web, desktop, mobile, and Codex CLI surfaces.
Repository:openai/codex
OpenAI Codex CLI: Local Terminal Coding Agent Built in Rust
Codex CLI is an open-source, lightweight coding agent developed by OpenAI that runs locally in your terminal. Written in Rust, it connects directly to ChatGPT plans or standard API keys.
121.7kstarsRust
Link:Anomaly
OpenCode: Open-source AI coding agent
OpenCode is an open-source AI coding agent running across terminals, IDEs, and desktop environments, offering integration with over 75 model providers.
Link:ecc.tools
ECC Tools: Open Harness System and GitHub App for Coding Agents
ECC Tools provides a cross-platform harness system for coding assistants such as Claude Code, Codex, Cursor, and OpenCode. It combines an open-source toolkit, an Opus 4.6-powered security auditor called AgentShield, and a GitHub App that transforms repository history into reusable skills.
Article:learn.chatgpt.com
Building Agent Skills for ChatGPT and Codex: Architecture and Best Practices
OpenAI details the implementation of agent skills for ChatGPT and Codex based on the open agent skills standard. Skills package instructions, references, and optional scripts into directory structures centered around a SKILL.md specification.
Let's talk about your project
Locations
- Mattersburg
Johann Nepomuk Bergerstraße 7/2/14
7210 Mattersburg, Austria - Vienna
Ungargasse 64-66/3/404
1030 Wien, Austria
Parts of this content were created with the assistance of AI.