Loading…

Loading…

Command Palette

Search for a command to run...

LearnOSDiscover

Sign in Sign up

Back to listing

Syllabus

Agent Harnesses: Anatomy, Build, Frameworks, Benchmarks— full course outline from your library.

Agent Harnesses: Anatomy, Build, Frameworks, Benchmarks

00 · Orientation

Agent Harnesses — Syllabus & How to Use This Course

01 · What a Harness Is

Model, Harness, Framework, Benchmark — and Why the Harness Moves the Score

The 10-Point Anatomy Rubric

02 · Teardowns: The Popular Harnesses

Claude Code — Teardown

Codex CLI — Teardown

OpenCode — Teardown

Pi — The Minimalist Counter-Argument

The Rest of the Field, Side by Side

03 · Build tinyharness I: The Core

Setup and the Provider Layer on OpenRouter

Tools and the Agent-Computer Interface

Context Engineering

04 · Build tinyharness II: Production-Grade

Sandboxing and Permissions

Sessions, Checkpoints, and Tracing

Extensibility: Hooks, Skills, and MCP

Subagents and Orchestration

Long-Running Agents

05 · Agent Frameworks

Raw API vs Framework vs Harness-as-Library

LangGraph and LangChain v1

The OpenAI Agents SDK

The Claude Agent SDK: a Harness as a Library

Google ADK, Pydantic AI, smolagents — and the Rest

06 · The Leading Benchmarks

How an Agent Benchmark Works

Coding I: The SWE-bench Lineage and Benchmark Rot

Coding II: Terminal-Bench and Harbor

Computer Use, Web Browsing and Tool Use

Economic Value and Long Horizons

07 · Harnesses for Evaluation & RL

Run tinyharness on Harbor

Build Your Own Eval

From Eval Harness to RL Environment

08 · Capstone

Capstone — Ship tinyharness on a Terminal-Bench Slice

Sign up to open