True enforcement
Hooks block unauthorized operations before they can occur. Your policy is enforced in real time. Agents can request their human's permission to bypass a guardrail when necessary.
Guardrails forcoding agents
File and command entitlement policy for you and your team's coding agents.
Instantaneous Install. No config files to maintain.
Install
curl -fsSL cordon.sh/install.sh | shOpen source. Free for individual developers.
Works with all of
tools you expect:
Claude Code
Cursor
GitHub Copilot
Gemini
OpenCode
OpenAI Codex
Claude Code
Cursor
GitHub Copilot
Gemini
OpenCode
OpenAI Codex
unsafe agent
Agent Entitlements
Agent entitlement policies for your engineering team
Define what coding agents can't do. Enforce it across every developer and every agent platform. See everything in one dashboard.
Audit everything
Every agent action is now immutably logged with attribution. Security and vulnerability analysis made easier. Incident reviews and token usage analysis becomes straightforward.
Team-wide policies
Define entitlements on-the-fly then sync and apply them immediately across developers, repos, and agent platforms. Stop agents from cheating the tests or wiping your production database.
Out of the box guardrails
$ cd ~/my-repository
$ cordon init
Add the standard guardrails?
These include sensible defaults for common footguns like destructive
commands ('git reset --hard', 'rm -rf') and credential file exposure (.env*).
Add standard guardrails? [Y/n]: Y
added 20 guardrail(s).
$ cordon command add 'supabase db reset --linked'How it works
Define and enforce
Define zones on files, folders, or patterns you want to protect.
Agents are blocked from writing to zoned files. Passes grant temporary access when needed.
Policy distribution
Ship policy instantly
Instantly distribute and manage entitlement policy for your teams.
Keep policy behavior consistent across local machines, agents, and repositories.
Get set up in minutes
Grab the latest release or build the code yourself.