Starting on its own
No human types - a message, a timer, an event starts the agent. Which history belongs to it, where does the answer go, and who applies the brake?
Why this is hard
A terminal agent has a built-in fence: the human in front of it. As soon as the agent answers on Telegram, wakes up at three in the morning via cron or reacts to a webhook, that fence is gone. Then the harness has to know three things itself: which history belongs to this inbox (the user, the thread, the channel?), where the answer goes, and when to stop if nobody can press Ctrl-C.
That is why triggers hang on two other layers. On the persistence of the message history, because an agent that starts on its own cannot keep its history in process memory - between two messages the process may not even exist. And on the budget of the guardrails, because the cap is now the only brake.
From Pro: the matrix per challenge, all eleven topics with citations, and the architecture portrait of every harness. Freely visible are the six short portraits:
- OpenClaw - The multi-channel personal agent - a half-million-line estate of gateway, channels and tooling whose loop core shares its lineage with Pi.
- Hermes - Nous Research's Python harness - deterministic loop, a skill system, and an error path that escalates in stages instead of just giving up.
- Pi - The coding agent with a deliberately small core - every layer hand-built and kept readable, from the provider package to the session file.
- HybridClaw - The enterprise runtime - sandboxed execution, approvals, and credentials that travel through context as references instead of plain text.
- DeepSeek Harness - The plugin radical - “everything is a plugin” is meant literally here: providers, tools and the loop itself are modules on the Cordis core.
- NanoClaw - The minimalist - small enough to understand in full: the model layer moves behind the Claude Agent SDK, isolation into a container.