Quick Start
Install ArmorClaude and see it enforce tool calls in under 2 minutes
Quick Start
This guide installs ArmorClaude 0.2.22. It uses the production backend and the armoriq CLI. Sign in to enable signed intent tokens and dashboard telemetry.
Prerequisites
- Claude Code 2.x or Claude Desktop
- Node.js 20.19 or newer, npm, Git, and
curl - macOS, Linux, or Windows with WSL or Git Bash
Install
curl -fsSL https://armoriq.ai/install_armorclaude.sh | bashcurl -fsSL https://armoriq.ai/install_armorclaude.sh | bashRun from WSL or Git Bash, not PowerShell/CMD.
The installer adds the armoriq marketplace, installs armorclaude@armoriq, installs @armoriq/sdk, and offers browser sign-in.
When prompted "Connect your ArmorIQ account now?", press Y to sign in through your browser. Declining keeps the plugin installed in monitor mode. Connect later with armoriq login --product armorclaude.
Manual install (without the curl script):
claude plugin marketplace add armoriq/armorClaude
claude plugin install armorclaude@armoriqVerify Installation
After the installer finishes, verify everything is set up:
claude plugin listYou should see:
Installed plugins:
armorclaude@armoriq
Version: 0.2.22
Scope: user
Status: enabledVerify the login:
armoriq whoamiTo connect or sign in again:
armoriq login --product armorclaudeThe browser approval saves the armorclaude profile for the selected backend. See Configuration for the profile and data-directory details.
Try It
Open Claude Code in any project:
mkdir -p /tmp/demo && cd /tmp/demo
echo "# My Project" > README.md
claudePrompt:
Create notes.md with one line describing this project.What happens:
- Claude registers an intent plan that includes
Write. - ArmorClaude records the plan and requests an intent token when you are signed in.
- Claude writes
notes.mdif the active policy permits it.
Read tools such as Read, Grep, and Glob do not require a plan, but still pass through policy checks. With the fresh-install all-allow policy, intent is recorded without blocking missing plans or drift. This exemption requires default allow and only permit statements; an empty default-deny policy does not qualify. Add a deny rule in the next step to test enforcement.
Block Something
A signed-in ArmorClaude follows your workspace's mode, which starts as Quiet. In Quiet the call below runs and is listed on the Monitor page instead of being blocked. To see the block in the chat, a workspace admin sets Tool calls in the chat to Block there, then you start a new Claude session. See Workspace Mode.
Set a policy rule using the /armorclaude:armor command. Replace the example statement ID below with the ID shown by policy list:
/armorclaude:armor policy add deny WebFetchArmorClaude stages the change and shows a diff. Confirm it:
/armorclaude:armor yesNow try:
Fetch https://example.comResult:
ArmorClaude policy forbid: policy1The tool call is blocked before execution. Claude sees the denial and tells the user.
To remove the rule:
/armorclaude:armor policy remove policy1
/armorclaude:armor yesWhat You Just Saw
| Feature | What happened |
|---|---|
| Intent plan | Claude declared the planned write |
| Policy enforcement | The active policy decided whether the call could run |
| Policy rule | deny WebFetch blocked the fetch call |
| Policy management | Rules created and deleted from the chat prompt |
Next Steps
- Policy Rules - rule effects, data classification, policy IR format
- Command Reference - complete
/armorclaude:armorcommand list - MCP Tools - intent registration, policy read, trust operations
- Policy Profiles - save, switch, and share policies
- Plan Mode - use with Claude's built-in plan mode
- Configuration - sign-in, environment variables, plugin settings
- Concepts - how intent enforcement works under the hood