Skip to Content

Model & Tool Ecosystem

Model Pool: SOTA Models Continuously Onboarded

  • Mainstream and SOTA large language models from multiple vendors are integrated; the model selector's real-time display is the source of truth for availability
  • Models are tiered into "Mainstream" and "SOTA"; SOTA models are clearly labeled in the selector and consume more units
  • Some models are restricted by plan or region; anything not shown in the selector is currently unavailable
  • Different nodes can be configured with different models (node-level model configuration): mainstream models for simple tasks, SOTA for core reasoning — the most effective lever for controlling usage

Account & Community MCP Tools

Five MCP tools deeply integrated with Longbridge's business, enabling Agents to serve users based on real account data:

MCP ToolCapability
Get Longbridge community user detailsSearch community users and their details by keyword, consistent with in-App search
Get detailed account informationThe user's current cash and stock positions (cost, share count, etc.)
Get P&L trend chartP&L trends by currency, time range, and time granularity
Get P&L detail dataP&L for a given security by time and stock code
Get P&L analysis summaryP&L summary by currency and time range

Prerequisite: calls involving asset data automatically trigger trading password verification (see Security Verification & Localization); data is returned only after verification passes.

Compliance reminder: account data may only be used for neutral display and statistical analysis — it must not be used to generate personalized buy/sell advice. See No Investment Advice and Staying Within the Execution-Only Positioning.

Time Zone Optimization for Market Data / Fundamentals Tools

  • Some tools' start_date / end_date parameter descriptions now include US ET and Hong Kong HKT time zone notes
  • All returned time fields are unified to ISO8601 format with UTC+0 time zone, with a timestamp of the API call time attached
  • Result: historical candlestick and historical financial report queries are more precise, and the model can determine the current trading session from the API time and overnight trading time points

What this means for builders: prompts for market-data Agents no longer need to labor over explaining time zones — tool returns carry explicit time zone information, reducing "yesterday/today" style temporal ambiguity errors.