@t2000/engine

The agent engine for conversational finance on Sui. AISDKEngine orchestrates LLM conversations, 26 financial tools, user confirmations, and MCP integrations into a single async-generator loop. Powers Audric.

Not a chatbot. A financial agent. Four systems work together to understand the user's money (Memory), reason about decisions (Reasoning Engine), act through 26 financial tools in one conversation (Agent Harness), and remember what it told them (AdviceLog). Every action it triggers still waits on user confirmation.

Install

npm install @t2000/engine @t2000/sdk

Requires Node.js 18+ · TypeScript 5+ recommended.

Quick start

import { AISDKEngine, getDefaultTools } from '@t2000/engine';
import { T2000 } from '@t2000/sdk';

const agent = await T2000.create();

const engine = new AISDKEngine({
  anthropicApiKey: process.env.ANTHROPIC_API_KEY,
  agent,
  tools: getDefaultTools(),
});

for await (const event of engine.submitMessage('What is my balance?')) {
  switch (event.type) {
    case 'text_delta':
      process.stdout.write(event.text);
      break;
    case 'tool_start':
      console.log(`\n[calling ${event.toolName}]`);
      break;
    case 'pending_action':
      // Write tool needs approval — client executes, then calls engine.resumeWithToolResult()
      break;
  }
}

For custom LLM providers or gateway routing, pass a pre-built LanguageModel via modelInstance instead of anthropicApiKey.

The 4 systems

| System | One-line | |---|---| | 🎛️ Agent Harness | 26 tools (18 read + 8 write), one agent. Parallel reads via the AI SDK step model; serial writes via a needsApproval round-trip. | | ⚡ Reasoning Engine | Thinks before it acts. Adaptive thinking effort, 12 guards across 3 priority tiers (Safety > Financial > UX), preflight validation, prompt caching. | | 🧠 Memory | Knows your finances. Vector-search-backed MemoryStore (recall + write) injected by the host; engine assembles a 4-layer system prompt with prepareStep (base → <memory_recall> → skill recipe → conversation). | | 📓 AdviceLog | Remembers what it told you. Host-side log (record_advice); last 30 days hydrate every turn so the chat never contradicts itself. |

The engine package owns Agent Harness and Reasoning Engine, plus the MemoryStore interface. Vector backends (e.g. MemWal) and the AdviceLog model live host-side.

Full reference

Event types, tool surface, USD-aware permission resolver, stream checkpoint resume, memory injection, MCP client + server adapters, host contract → developers.t2000.ai/agent-engine

License

MIT — see LICENSE.