[01] Harness · required/ What thinks

Harness for AI agents

The loop that runs your agent. Take one off the shelf, or skip it and write the loop yourself.

12 options tracked · 10 open source · 8 self-hostable · required in every stack

· All 12 options/ Compare
OptionWhat it doesLicenceSelf-host
Claude CodeAnthropic's coding agent — terminal, IDE and SDK, with hooks, skills and MCP built in.ProprietaryNo
Claude Agent SDKThe Claude Code loop as a library, for agents that are not about code.ProprietaryNo
Hermes AgentSelf-improving agent from Nous Research — writes its own skills from experience and carries them forward.Open sourceYes
OpenClawSelf-hosted gateway putting an agent in Slack, Telegram, WhatsApp, Signal, Discord and iMessage at once.Open sourceYes
Prime AgentPrime Intellect's recursive-language-model harness — a persistent IPython kernel is the only tool, sub-agents are function calls.Open sourceYes
DeepSeek HarnessEverything is a plugin — the model adapter, the tool registry, the sandbox and the agent loop itself are all swappable.Open sourceYes
OpenAI Agents SDKLightweight loop with handoffs, guardrails and sessions.Open sourceNo
Codex CLIOpenAI's terminal coding agent.Open sourceNo
OpenCodeOpen-source terminal coding agent, provider-agnostic.Open sourceYes
GooseBlock's open-source on-machine agent with an extension system.Open sourceYes
LettaStateful agent server where memory is the primary abstraction.Open sourceYes
Roll your own loopNo harness. A while-loop, a model call and your own tool dispatch.Open sourceYes
· Head to head/ 55 comparisons
· Common questions/ FAQ

What is the harness layer of an AI agent?

The loop that runs your agent. Take one off the shelf, or skip it and write the loop yourself. Every stack needs one — it is not optional.

How many harness options are there?

This registry tracks 12. 10 are open source and 8 can run on your own infrastructure.

Which harness option should I choose?

It depends on constraints rather than preference: whether you must self-host, whether the budget allows a hosted service, and which language your team writes. Describe what you are building and the advisor fills this layer along with the other 9.

· The other 9 layers/ Keep going
· For agents/ This page, machine-readable

Every page here answers to Accept: text/markdown and returns the same content at roughly a tenth the tokens. No separate site, no toggle — same URL.

curl -s -H "Accept: text/markdown" https://newagent.build/layers/harness