Developer Documentation

Install the official client package to start instrumenting your agent runs:

npm install @ai-sage/client

Initialize the client with your agent ID and environment API key:

import { AISage } from "@ai-sage/client";

const sage = new AISage({
  agentId: "refund-agent-prod",
  apiKey: process.env.AI_SAGE_API_KEY,
});

await sage.guard(async (ctx) => {
  const reply = await agent.run(input);
  ctx.assert.cost.below(0.40);
  return reply;
});

Define runtime policies directly in your dashboard or config files:

• max_cost_per_run: Automatically throttle or halt agents exceeding budget caps.
• allowed_tools: Restrict tool execution to explicit white lists.
• min_confidence: Trigger automated human fallbacks when agent confidence drops below your threshold.

Connect your live n8n AI workflow nodes to AI Sage to monitor node execution health, token spend, and trigger automated pause actions during runaway loops.

API Reference