Design Centers

Design Centers: OpenClaw, Hermes, and Philont

This is a comparison of emphasis, not a feature scorecard. These projects overlap substantially and change quickly; a checkmark table becomes inaccurate faster than it becomes useful.

Project	Primary design center
OpenClaw	Broad integrations and a practical personal-agent ecosystem
Hermes Agent	Polished task execution, channels, reusable skills, and experience-driven improvement
Philont	Persistent pursuits and a runtime that makes authorization, execution evidence, recovery, and learning governance explicit

Philont's distinct question is:

Can an auditable, persistent runtime make smaller models dependable enough for long-running autonomous work?

That question produces the project's characteristic mechanisms: compass.md, persistent pursuits, plan and reasoning state machines, execution-ledger checks, scoped grants, resumable authorization, and a skill lifecycle with validation, demotion, revision, and repair.

This is a positioning statement, not a claim that the result has already been proven. Evaluation Status tracks the evidence still needed.

中文

这是一份“设计重心”比较，不是功能打分表。三个项目的能力大量重叠，也都在快速变化；勾叉表往往还没产生价值就已经过时。

项目	主要设计重心
OpenClaw	广泛集成与实用的个人 Agent 生态
Hermes Agent	成熟的任务执行、渠道、可复用技能与经验驱动改进
Philont	长期 pursuit，以及把授权、执行证据、恢复和学习治理显式化的 runtime

Philont 独特的问题是：

一个可审计、可持续的 runtime，能否让较小模型可靠到足以承担长期自主工作？

这是定位和研究命题，不是已经被证明的结果。参见评测状态。