CTRLRun
An open-source safety layer that checks every AI agent action before it runs
Emerged finds new products worth a look, daily — what’s new today
DashClaw
Approval gates and an audit ledger for coding agents left running unattended
It lets you leave Claude Code or Codex running overnight while a human still signs off on anything destructive.
DashClaw sits between unattended coding agents and their tools. It intercepts risky calls, applies policy, routes the ones that need a person to an approval queue, and records every decision in an auditable ledger. It works with Claude Code, Codex and others, runs locally with npx dashclaw up, ships npm and pip SDKs, and has a hosted trial. MIT licensed, 302 stars, over 3,000 commits; notably, day-to-day development is done by an agent under a published human charter.
More like this
no_human
Open-source agent that takes a ticket to a reviewed pull request, with proof
Shield KYA
A local ledger of every tool call your coding agents make
AgentRun
Open-source DSL that freezes repeated agent work into typed, inspectable steps
OpenBot
Self-hosted AI coworkers, each with its own computer and a policy gateway
Imza
A tiny native terminal agent where you co-write the spec before any code