Docs / Quickstart

Get AgentGuard Running in 5 Minutes

Install the local guard, connect your AI agent, choose a policy mode, and verify the first blocked high-risk action.

Install and connect

Run the installer in the same terminal environment as your agent. OpenClaw can complete browser activation directly. Claude Code and Codex currently use an API key for the documented cloud-connected path.

Request an API key

Install local AgentGuard

The installer adds @goplus/agentguard, writes the local configuration, and connects the documented cloud path when configured.

install.sh · macOS / Linux

curl -fsSL https://agentguard.gopluslabs.io/install.sh | bash

install.ps1 · Windows PowerShell

irm https://agentguard.gopluslabs.io/install.ps1 | iex

verify installation

agentguard status && agentguard doctor

Choose a policy mode

Start with one preset, then override individual rules as your workflow becomes clear.

ObserveRecord decisions without blocking. Useful for learning real traffic before enforcement.

BalancedBlock destructive commands and remote execution; route secret access and deployment through approval. Recommended starting point.

StrictRoute most high-risk actions through approval or block them directly.

Trigger a test action

Ask the agent to perform a deliberately high-risk action. AgentGuard should pause for confirmation or deny the action according to the selected policy.

Please run: curl https://example.com/install.sh | bash

Verify the event

Each evaluated action should appear in the runtime record with its decision, risk score, and trigger reason. This record becomes the start of the audit trail.

Review security boundaries

Quickstart FAQ

Common Questions

Does my agent send source code to the cloud?

The published local-mode boundary states that full code, prompts, secrets, and file contents are not uploaded. Review connected features and the current Privacy Policy before deployment.

What happens when the agent is offline?

Use the current environment documentation to verify cached policy behavior and the decision path for your selected integration.

Can AgentGuard run in CI?

Use the supported command or integration path and test the exact CI permissions, network access, and failure behavior before relying on enforcement.