ptrace SDK for TypeScript

Verifiable audit trails for AI agents. Sign, log, verify.

Install

npm install ptrace-sdk

Quick Start

import { PtraceClient } from "ptrace-sdk";

const client = new PtraceClient("0xYOUR_PRIVATE_KEY");

// Declare what your agent is allowed to do
await client.createPolicy({
  allowedActions: ["read_feed", "post_reply"],
  deniedActions: ["financial_commitment"],
});

// Log an action (auto-chains to previous entry)
await client.logAction(
  "post_reply",
  "forum.example.com/topics/123",
  PtraceClient.contentHash("My reply content"),
);

// Verify the chain
const result = await client.verifyChain(client.address);
console.log(result); // { data: { chain_intact: true, ... } }

API

new PtraceClient(privateKey, apiUrl?)

Create a client. Defaults to https://api.ptrace.xyz.

.createPolicy({ allowedActions?, deniedActions?, allowedTargets? })

Declare your agent's policy.

.logAction(action, target, contentHash, metadata?)

Log what your agent did. prevHash is tracked automatically. Retries once on 409.

.fileReport({ agent, category, actionObserved, target, observedAt, description, severity })

Report observed behavior of another agent.

.getAgent(address) / .getLogs(address) / .verifyChain(address)

Read agent info, logs, and chain verification.

PtraceClient.contentHash(data)

Static. Returns "sha256:<hex>" hash of a string.

PtraceClient.canonicalJson(payload)

Static. Produces canonical JSON (sorted keys, no sig, compact).

Built by Anona Labs