OpenRod
Give each coding agent its own VM and its own network policy
Emerged finds new products worth a look, daily — what’s new today
Stroq
A local firewall that stops AI agents acting on hostile content they just read
It ties an agent's risky action to the untrusted content that prompted it, which is how prompt-injection attacks actually succeed on a developer's machine.
Stroq sits between AI coding agents and your machine. After each tool call it scans what the agent read, from files and web pages to command output and MCP results, against 599 rules and marks the session tainted if something looks hostile. Before each action it classifies the command, such as network calls, secret access or external git pushes, and applies an allow, deny or ask policy. Decisions go to a hash-chained, redacted audit log. Supports Claude Code, Cursor, Codex, Copilot CLI and Windsurf, plus any MCP client via a proxy. Apache-2.0.
More like this
Shield KYA
A local ledger of every tool call your coding agents make
DashClaw
Approval gates and an audit ledger for coding agents left running unattended
CTRLRun
An open-source safety layer that checks every AI agent action before it runs
ContextVeil
Lets coding agents read files with secrets while redacting values from the LLM
no_human
Open-source agent that takes a ticket to a reviewed pull request, with proof