Observed arrival · 2026-10-03
RubyClaw: an agent harness that writes and checks its own tools
RubyClaw is an open-source Ruby harness for AI agents, designed to create and verify tools, manage persistent tasks, and require approval before restricted actions.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Ruby developers building AI agent workflows
- Worth noticing
- The page says unmatched policy actions default to ask, and the agent cannot grant its own approvals.
Field notes
The documented tool workflow checks generated code in a throwaway child process and exercises it with real arguments before adding it to the live tool set. Persistent work is represented as task records with eight states, an append-only event log, and artifacts tied to tasks. The policy system distinguishes automatic, approval-required, blocked, and human-only actions; the page says an unmatched action requires approval.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue