FinSight
Automate multi-modal financial deep research and report generation with code-first agents.
One ticker, one click, one publication-ready financial report.
FinSight is a multi-agent system that automates the entire financial research pipeline from data collection to final report generation. It suits quantitative researchers, financial analysts, and developers looking to produce professional equity and macro reports automatically.
What it is
The system combines a programmable Code Agent with Variable Memory (CAVM) and a VLM-powered chart critique loop to write 20,000+ word financial reports. It fetches data across A-share, HK, and US equities via yfinance, along with macroeconomic series via FRED.
Who it's for
It suits financial professionals and engineers who need deep, programmatic research backed by verifiable code and data citations. It does not suit casual users looking for simple stock tickers without configuring LLM API keys and Python environments.
Key features
Code Agent Variable Memory
Agents operate in a unified variable space, executing Python code to manipulate data and tools dynamically.
VLM Chart Critique Loop
Visual agents correct missing legends, incorrect scales, and low information density automatically.
Multi-Market Data Coverage
Access real-time stock profiles, financial statements, and macro indicators across US, A-share, and HK markets.
Evidence-Traced Analysis
Every conclusion is derived from a transparent chain of analysis with strict citations linking back to original sources.
Up and running in 3 steps
Our honest verdict
What's great
- Generates 20,000+ word publication-ready financial reports with professional charts
- Supports multi-market equity and macroeconomic data via yfinance and FRED
- Implements Code Agent with Variable Memory for reproducible, transparent analysis
- Includes both a full CLI pipeline and an interactive web UI with checkpointing
Where it falls short
- Requires external API keys for your LLM stack, VLM, embeddings, and search providers
- Running long-form research tasks and multiple model calls incurs significant API token costs
- Requires Python 3.10+, Pandoc, and optional Node.js to build the web interface from source
Pricing & license
Free under its open-source licence, with no paid hosted version provided. The only cost is running it yourself and paying for the underlying LLM, VLM, and search API tokens.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
What data sources does FinSight support?
It supports A-share, HK, and US markets via yfinance, macro indicators via FRED, and web search integration.
Can I output reports to formats other than Markdown?
Yes, using Pandoc you can export polished DOCX and PDF documents directly from the pipeline.
How does it avoid ugly charts?
It features a built-in visual language model (VLM) feedback loop that critiques and corrects charts automatically.
