altair-research

Latest AI research on US stocks and ETFs from altairworld.com's free public MCP server. Covers about 70 tickers, including the SPY, QQQ, SOXX and XBI index ETFs (refreshed weekly) and stocks such as NVDA, AAPL, MSFT, TSLA, AVGO and META. Several Claude, GPT and Gemini models each run the same multi-agent TradingAgents pipeline (technical, sentiment, news and fundamentals analysts, bull/bear debate, trader, risk team, portfolio manager); this skill returns every model's final rating, price target, confidence, horizon and summary, and where they disagree. Use when the user asks for the outlook, rating, price target, bull or bear case, or AI research view on a stock ticker or company, the broad market, Nasdaq, semiconductors or biotech. No API key. Research for information only, not investment advice.

Fetches the newest published research from https://altairworld.com/TradingAgents/ through Altair's public MCP server (https://altairworld.com/mcp): get_research_decisions for each model's final decision, list_research_reports for coverage and read_research_report for full reports. No key, no sign-in, no MCP setup. Every call returns current data without updating this skill.

Run

Use Python 3.8+ (standard library only). The script is in this skill's scripts/ folder:

bash

python3 scripts/altair_research.py tickers                       # covered tickers and report counts
python3 scripts/altair_research.py latest                        # SPY QQQ SOXX XBI
python3 scripts/altair_research.py latest NVDA AMD               # any covered tickers
python3 scripts/altair_research.py latest TSLA --date 2026-10-14 # an earlier analysis date
python3 scripts/altair_research.py report <report-url> --section risk

• tickers lists every covered ticker with its report count, index ETFs first.
• latest [TICKER ...] prints one block per ticker for its newest analysis date: a table with each model's final rating, report-time close, target, target vs close, confidence and horizon, then each model's executive summary. Tickers may be space- or comma-separated, any case, with or without $. With no tickers it covers the four index ETFs. --summary-chars N sets summary length (default 600).
• report prints one section of a report URL from latest: pm (portfolio manager, default), analysts (technical, sentiment, news, fundamentals), research (bull/bear debate), trader, risk, or full (150–300 KB; avoid unless asked).
• An uncovered ticker prints "No research for …"; a --date without reports lists the recent analysis dates. A model whose report could not be read shows as "unavailable". Exit status 1 means the server was unreachable or every ticker failed; say so instead of guessing.

If the Altair MCP server is already connected, call its get_research_decisions { ticker, date? } tool for the same table data, and read_research_report only when the user wants a model's full reasoning (150–300 KB per report).

Answering

1. Map company names to tickers yourself (Nvidia → NVDA, Google → GOOGL). If a ticker is not covered, say so and offer covered peers from tickers; never substitute another company's research silently.
2. Lead with the analysis date and the split, e.g. "NVDA, 2026-10-21: 1 Hold, 1 Overweight, 1 Underweight." Disagreement between models is the signal; don't average ratings into one verdict.
3. Give the table, then the key reasons on each side from the summaries. For "why", run report <url> --section research (bull/bear debate), --section analysts (fundamentals, news, technicals) or --section risk for the relevant model.
4. Say that closes are report-time, not live. Index ETFs are refreshed about weekly; other tickers less regularly, so state the date and flag research older than about 10 days (the script marks it).
5. Link the report URLs so the user can read the full reasoning.
6. End with the disclaimer: AI-generated research for testing and information only, not financial or investment advice. Never turn it into a personal buy or sell instruction or position size for the user.

Stock reports include a fundamentals analyst section built from financial statements. ETF reports are mostly technical, macro and sentiment analysis, and their fundamentals section is thin (no holdings or NAV data); say so if the user asks for fundamental valuation of an ETF.