Call the SDK

Install @limetry/sdk, build an ActionIntent, and evaluate it against a self-run Limetry node.

@limetry/sdk is open source. Point baseUrl at your evaluation server (http://localhost:3810 or your origin). Use LIMETRY_BEARER_TOKEN unless you issue your own API tokens on that node.

Install the SDK

npm install @limetry/sdk

Until the first npm release is live, install from the source checkout (see INSTALL.md).

Build an ActionIntent

import type { ActionIntent, ActionPolicy } from "@limetry/sdk"
import { createSlimActionPolicy, redactActionIntent } from "@limetry/sdk"

const policy = createSlimActionPolicy({
  agentId: "demo_agent",
  allowedActionTypes: ["http_get"],
  deniedActionTypes: ["http_post", "deploy"],
  blockedResourcePatterns: ["https://prod.example.com/*"],
  maxCostMinor: 100,
  auditMode: "minimal",
})

const intent: ActionIntent = {
  intent_id: crypto.randomUUID(),
  policy_id: policy.policy_id,
  agent_id: "demo_agent",
  action_type: "http_post",
  resource: "https://prod.example.com/api/charge",
  issued_at: new Date().toISOString(),
}

/**
 * Evaluate with the full intent. Use redact helpers for logs / client-side telemetry.
 */
const safeForLogs = redactActionIntent(intent)

Send only fields needed for policy. Prefer auditMode: "minimal" unless you need forensics. See Minimize audit data.

Evaluate over HTTP

import { createRemoteEngine } from "@limetry/sdk"

const engine = createRemoteEngine({
  baseUrl: "http://localhost:3810",
  apiKey: process.env.LIMETRY_BEARER_TOKEN,
})

const decision = await engine.evaluateAction(intent)

createRemoteEngine() also reads LIMETRY_API_KEY / LIMETRY_BEARER_TOKEN and LIMETRY_BASE_URL.

Set LIMETRY_BASE_URL to your evaluation server origin when it is not localhost.

Handle evaluate errors

Error	When
`PolicyViolationError`	Evaluate returned deny
`ApprovalRequiredError`	Evaluate returned `approval_required`
`NetworkError`	Server unreachable or timed out

Continue

Evaluate your first intent

Self-host and evaluate a sample intent.

Minimize audit data

Audit modes, secret scrubbing, and retention.

Self-run the server

Deploy the evaluation server.