0. 30-second summary

• In March–April 2026, Anthropic and OpenAI raised "harness" as an official topic of discussion about three weeks apart.
  • Anthropic — 2026.3.24, "Designing harnesses for long-running applications" (Prithvi Rajasekaran)
  • OpenAI — 2026.4.15, Agents SDK 2.0 — long-horizon harness, sandbox, subagents
• MCP, skills, and CLIs are being pushed as standardized and open — but for the harness layer, both vendors are signaling "design it inside our SDK."
• At the same time, infra costs for both PCs and local inference are exploding. Even though Google's TurboQuant (2026.3.25) cuts memory by ~6×, the Jevons Paradox means total demand goes up, not down.
• We have to pick one of two paths — play on top of a vendor-locked harness, or design a hybrid native harness ourselves.

1. Why is everyone suddenly talking about "the harness"?

A "harness" is originally horse tack — gear that converts a horse's raw power into controllable direction. In AI, the term now refers to the entire outer-layer system that turns the LLM (the horse) into a controllable worker.

"Claude Code serves as the agentic harness around Claude: it provides the tools, context management, and execution environment that turn a language model into a capable coding agent."

— Anthropic official docs (Memorizer: How Claude Code Works — Harness, AI Agent, Agentic Loop)

The six core components are:

flowchart LR
    Model["LLM<br/>(the horse = inference engine)"]
    subgraph Harness["Harness (the tack)"]
        Tools["Tools<br/>Bash / Read / Write"]
        Perm["Permissions &<br/>approval gates"]
        Sandbox["Sandbox"]
        Session["Session &<br/>memory"]
        Context["Context window<br/>management"]
        Ext["Extensions:<br/>MCP / Skills / Hooks"]
    end
    Model --> Harness
    Harness --> Agent["AI Agent<br/>(judges & acts on its own)"]

Looking at the last two years, the standardization wave moved from model → tools → MCP → skills → CLI. Then in spring 2026, both vendors simultaneously tried to grab the reins back on one specific layer: the harness.

2. Anthropic's harness — "Start simple, add complexity only when needed"

Primary source: <https://www.anthropic.com/engineering/harness-design-long-running-apps> (2026.3.24)

Anthropic published its design principles for harnesses targeting long-running apps (jobs that span hours or days).

2.1 Design philosophy

"Find the simplest solution possible, and only increase complexity when needed."