AgentGuard
io.github.MerchantGuard/agentguard
Local spend caps, kill-switch blocking and Ed25519-signed receipts for AI agent tool calls.
Install surfaces
How AgentGuard runs, per the official MCP registry entry.
- npm
npx @agentguard-run/[email protected]
Governed execution
How KLA gates calls to AgentGuard once it is connected.
A published policy defines which AgentGuard actions an agent may take and under what conditions. Each call passes through a policy gate at the tool-call boundary. Permitted actions execute, with the policy basis recorded. Flagged actions execute and are sampled for scheduled human review. High-stakes actions pause and route to a named reviewer on the Decision Desk. The call, the policy decision, and any approval are sealed as an evidence record.
KLA also governs the agents and model providers you already run. See integrations.