Menu
Product / AgentGuard SaaS
EARLY ACCESSAgentGuard, hosted for you.
We run the platform. You connect your agent with our SDK and see its runs, controls and evidence in one place.
From sign-up to your first run
- 01
Create your account.
Sign up with your work email and confirm it with the code we send you.
- 02
Connect your agent.
Create a project, copy its keys and add the SDK.
JavaScript npm install @actaclad/agentguardPython pip install actaclad-agentguard - 03
See your first run.
Run your agent. Each run shows up in AgentGuard with its model calls, tool calls and control decisions.
On day one
One agent.
Every decision, connected.
See every run
AgentGuard captures each agent run as a trace: model calls, tool calls, latency and token usage.
Protect sensitive data
AgentGuard guardrails detect and redact sensitive data such as PII, secrets and tokens in prompts and outputs.
Set tool permissions
AgentGuard tool_permission sets each tool an agent can call to ALLOW, READ_ONLY, APPROVAL_REQUIRED or BLOCK, with a per-tool call cap.
Check every response
AgentGuard checks model responses with output guardrails such as schema validation, hallucination and toxic-content checks.
Apply your own policies
Custom policies apply keyword or regex rules to inputs and outputs, scoped to selected agents.
Keep the evidence
AgentGuard records each control decision as evidence.
Good to know
Need AgentGuard in your own cloud or on your own VMs, or a security review before you start? Talk to ActaClad →