NSE Multi-Agent Stock Research System
Four coordinated AI agents pull live NSE prices, indicators and news sentiment to produce actionable BUY/SELL/HOLD calls with entry, stop-loss and confidence.
- Source repo
- rooneyrulz/agentic-stock-research-system
- Stars
- ★ 125
- Last updated
- 2mo ago
- Primary language
- Python
- FA score
- 38/100 · Major gaps
At a glance
- Works with
- Portable with changesOpenAI API
- You'll need
- Typical use
- Retail investors following Indian equities who want technicals, volume and news sentiment combined in a single view.
- Main limitation
- Hard dependency on two paid external services (Bright Data scraping and OpenAI API); a scraping failure breaks the analysis chain.
- Maintenance
- Commits in the last 6 monthslast commit 2026-09-15
- Source review
- 38/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
This is a Streamlit research tool for Indian NSE-listed equities built around a Supervisor agent and four specialist agents: a Stock Finder Agent that screens for liquid, large- and mid-cap names by momentum, a Market Data Agent that collects real-time prices and volume and computes RSI, MACD and moving averages, a News Analyst Agent that scrapes recent financial news and classifies sentiment, and a Recommendation Agent that synthesizes everything into BUY/SELL/HOLD calls with entry ranges, stop-losses, target prices and confidence scores. Data acquisition goes through Bright Data's MCP server (the @brightdata/mcp npm package installed globally plus a Bright Data token), and reasoning uses an OpenAI API key; both credentials are entered in the app sidebar. You run it locally with streamlit run streamlit_app.py, open localhost:8501, pick a short-term, medium-term or general-market analysis type, and click Start Analysis. Output appears as interactive charts, formatted recommendation text and a CSV export, and the README frames the whole thing as educational and research use only.
After the user enters a Bright Data API token and an OpenAI API key in the Streamlit sidebar and chooses an analysis type (short-term 1-7 days, medium-term 1-4 weeks, or general market overview), the Supervisor agent orchestrates the pipeline. The Stock Finder Agent scans the NSE universe and filters out penny and illiquid stocks; the Market Data Agent calls the Bright Data MCP Server for real-time pricing and volume and derives RSI, MACD, moving averages and support/resistance levels; the News Analyst Agent scrapes recent financial news and announcements and labels them Positive/Negative/Neutral. The Recommendation Agent then merges these into a BUY/SELL/HOLD verdict with a suggested entry range, stop-loss, target price, risk-reward ratio and confidence level, rendered as charts and exportable CSV in the UI.
- Retail investors following Indian equities who want technicals, volume and news sentiment combined in a single view.
- Short-term (1-7 day) traders looking for concrete entry bands, stop-losses and target prices to compare against their own plan.
- Users who want a sector-wide overview of NSE opportunities via the General Market Analysis mode.
- Analysts who want to export results as CSV for further statistics or record-keeping.
- Developers studying Supervisor-plus-specialist agent orchestration with MCP-based data access.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Install / run
pip install -r requirements.txtnpm install -g @brightdata/mcppip install -r requirements.txt --force-reinstall
- Config files
.env.example.env- Runtime
- Python 3.8+
How do you install or deploy this agent?
You need Python 3.8+, Node.js/npm, a Bright Data account (grab the token under Zones → Web Unlocker) and an OpenAI API key starting with sk-.
git clone https://github.com/rooneyrulz/agentic-stock-research-system
cd nse-stock-research-system
pip install -r requirements.txt
cp .env.example .env
npm install -g @brightdata/mcpThen edit .env with both credentials. If installation breaks, the README suggests pip install -r requirements.txt --force-reinstall; for MCP problems, uninstall and reinstall the global npm package.
How do you use this agent?
Launch the app and open port 8501 in a browser:
streamlit run streamlit_app.pyEnter the Bright Data token and OpenAI key in the sidebar, select an analysis type, and click Start Analysis; wait while the agents fetch data, compute indicators and produce recommendations. Review the interactive charts, copy the formatted recommendation text, or download the CSV export. If keys fail, verify the Bright Data token has credits and the OpenAI key has quota; if Streamlit misbehaves, run streamlit cache clear.
What are this agent's strengths and limitations?
- Clean separation of concerns: four specialist agents each own one stage (screening, market data, news, recommendation) under a Supervisor, so any single stage can be swapped.
- Actionable output shape: every call carries an entry range, stop-loss, target, risk-reward ratio and confidence score rather than a vague bullish/bearish view.
- Uses Bright Data's MCP web-scraping server instead of a paid market-data subscription, and computes RSI/MACD/moving averages locally.
- Ships a Streamlit UI with interactive charts and CSV export, usable without writing code.
- README states risk disclaimers and educational scope, and lists troubleshooting commands for common failures.
- Hard dependency on two paid external services (Bright Data scraping and OpenAI API); a scraping failure breaks the analysis chain.
- Requires a globally installed Node package (@brightdata/mcp) alongside the Python stack, adding a second runtime.
- License information is inconsistent: the repository metadata says unknown while the README claims MIT, so verify before adopting.
- Scope is limited to NSE India and web-scraped data; there is no official market-data API or backtesting framework.
- Recommendation quality depends entirely on LLM reasoning, with no published accuracy tests or evaluation of past calls.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| NSE Multi-Agent Stock Research System This agent | 38 · Major gaps | — | ★ 125 | 2mo ago | Python | OpenAI API |
| PRISM-INSIGHT | 52 · Major gaps | CLIFree + model costs | ★ 812 | today | Python | ChatGPT · OpenAI API |
| QuantHarness: Price-Driven Multi-Agent LLMs for High-Frequency Trading | 38 · Major gaps | Web appFree + model costs | ★ 2.9k | 3mo ago | HTML | OpenAI API · Claude API |
| Yuxi – Self-Hosted Multi-Tenant Knowledge Agent Platform | 85 · Good | Self-hosted serviceFree + model costs | ★ 7.7k | 1d ago | Python | OpenAI API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows the system launches npx @brightdata/mcp over stdio and injects API_TOKEN; scope is largely limited to scraping and LLM calls, but there is no least-privilege constraint on tool use and no user confirmation step. API keys are entered as password fields in the Streamlit sidebar, yet no explicit persistence or log-redaction handling is visible. Dependencies are declared without upper bounds in requirements.txt, adding supply-chain uncertainty. No rollback/revocation mechanism exists. README and code disagree (README says OpenAI, code uses Groq), weakening attribution. Most items score 1; rollback is 0.
Structure is clear but main.py is truncated at analyze_stocks, so the full flow cannot be confirmed. Dependency versions diverge between pyproject and requirements (e.g., langchain-groq missing from requirements), a availability risk. Error handling and failure messages appear only as logs and sidebar validation, with no clear user-facing failure messaging. All score 1.
README clearly targets Indian NSE stocks with short/medium/general scenarios, giving adequate audience/scenario coverage (2). Capability boundaries (NSE-only, external API quota dependence) are not systematically stated; triggers rely on free-text queries with limited precision; environment requirements conflict (Python 3.8+ in README vs 3.12 in pyproject). Boundaries/trigger/environment score 1.
README has solid information architecture, install steps and troubleshooting (2/2), and concrete sample output (2). But naming is inconsistent (README says streamlit_app.py, file is app.py), version history is only a brief README list, LICENSE is referenced but not included in the packet, and maintenance responsibility/update path are unclear. Naming/versioning/license/maintenance score 1; known limitations only reach disclaimer level (1).
Output includes structured recommendations, CSV export and charts, giving good usability (2). However parsing relies on regex extraction from free text, robustness is questionable, marginal value is limited by a single market and external API dependence, and cost-benefit is moderate given dual paid Bright Data and Groq requirements. Marginal value and cost-benefit score 1.
Recommendations rest on live scraping and LLM reasoning, but there is no per-item source citation or fact/inference separation labeling; README sample output is a static example that cannot be traced; cross-source corroboration depends on a single Bright Data channel. All score 1.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- README and code disagree: README requires an OpenAI API key, but main.py actually uses ChatGroq/GROQ_API_KEY, which may mislead configuration.
- Dependency declarations are inconsistent and unbounded: requirements.txt omits langchain-groq, and pyproject requires Python>=3.12 while README says 3.8+, risking install failure.
- main.py is truncated at analyze_stocks, so the full workflow, error handling and output parsing cannot be confirmed.
- Recommendation parsing relies on regex extraction from free text and may produce errors or empty results when formats change.
- The system needs dual paid APIs (Bright Data and Groq); cost and quota depend on external services, and no rollback/revocation mechanism is visible.
- The LICENSE file is not included in the packet; only README references MIT, and license metadata is unknown.
FAQ
Is NSE Multi-Agent Stock Research System open source?
Is NSE Multi-Agent Stock Research System safe to use?
Is NSE Multi-Agent Stock Research System still maintained?
Who makes NSE Multi-Agent Stock Research System?
Thanks — we use this to improve the page.