Why it works
The OpenClaw lineage
Built from OpenClaw's ideas, not on its runtime — and where we diverged.
pal.fun's architecture comes from OpenClaw, the MIT-licensed open-source agent runtime. The core ideas in this section — identity as markdown, context as code, persona as prose — are OpenClaw's. We took those ideas and built our own runtime on them: pal.fun doesn't run OpenClaw's code, it runs on OpenClaw's thinking.
What we lifted
The load-bearing ones, almost verbatim:
- Identity as markdown. An agent is a set of human-readable files — who it is, what it knows, how it works — not a config blob. That's the whole premise of the glass box.
- Context as code. What the agent perceives each turn is assembled, deliberately, from those files plus live state — not dumped in wholesale. (Context engineering.)
- Persona via prose. Character is written, never configured. (Persona via prose.)
- Negative space. Half the craft is telling the model what not to do. (Negative space.)
If those feel like the spine of the whole product, that's because they are — and they're OpenClaw's spine first.
Where we diverged — on purpose
A homage that only flatters isn't worth much. The interesting part is where we deliberately departed, because that's where pal.fun's own shape comes from.
- Operator → normie. OpenClaw assumes a technically fluent operator — someone who lives in the files, edits them, runs the agent like a power tool. pal.fun assumes someone who will never open a terminal. So the files are read-only for the human, and the pal does its own curation by talking to you. Same transparency, opposite hands on the keyboard.
- Transparency as trust, not control. For an operator, seeing everything is about control — you're tuning a machine. For our user, seeing everything is about trust — you're getting to know a Someone. The glass box is the same window; what it's for is different.
- No continuous-agent vocabulary. OpenClaw thinks in long-running, self-scheduling processes. We dropped that wholesale: a pal is never on a cron, never "running in the background." Its proactivity comes from the relationship — the onboarding, the things it chose to remember — not from a daemon. Importing the always-on framing would have quietly turned a friend back into infrastructure.
Credit
The patterns are OpenClaw's; the decisions about who they're for — a person who wants a Someone, not a runtime — are ours. If you build agents, OpenClaw is worth your time. It's where a lot of this thinking gets worked out in the open, and you can read every line of it.