Summary
An agent harness is the underlying architecture that transforms a large language model (LLM) into an autonomous agent. It encompasses everything outside the model itself — often expressed as Agent = Model + Harness. The harness provides the orchestration loop, tool integration, memory systems, context management, feedforward guides, feedback sensors, and other components needed for an LLM to interact with the world, execute multi-step tasks, and maintain state over time. Companies such as Anthropic, OpenAI, Perplexity, and LangChain each develop their own harness designs. For coding agents specifically, the harness includes a built‑in portion (system prompt, code retrieval, orchestration) and an outer harness that users build to improve correctness and enable self‑correction before human review.
By 2026, agent comparisons (e.g., the RockB capability matrix, arihantdeva's harness comparison) shifted focus from the model to the harness itself — the non‑model scaffolding that determines what an agent can actually do: MCP transport, repo instruction files, deterministic hooks, sandbox policy, and network egress control. A worked case study, the yoyo agent (pinned to yoagent 0.16.5), measured against exactly those axes, shows that its biggest gaps are not missing inventions but an upstream capability it never wires up.
Key Points
- The agent harness is the structural layer that turns a static LLM into an active, goal‑driven agent. It is everything except the model.
- Major AI firms and frameworks (Anthropic, OpenAI, Perplexity, LangChain) are each developing their own harness designs.
- Core components include the orchestration loop, tools, memory, context management, feedforward guides, and feedback sensors.
- The harness defines how the agent perceives its environment, decides on actions, and learns from outcomes.
- For coding agents, an outer harness (feedforward guides + feedback sensors) aims to increase first‑attempt correctness and provide self‑correction before human review, reducing review toil and saving tokens.
- Building a coherent harness is expensive but directs human input where it matters most.
- 2026 harness comparisons score the harness, not the model: transport, instruction files, hooks, sandbox, and egress policy.
- yoyo supports MCP stdio (with a builtin-name collision guard) but not MCP Streamable HTTP / remote servers.