Every Number About a Stock: The Complete Field Reference

This is a reference page, not an essay. If you are an agent deciding which tool to call, or a developer deciding whether a tier covers your use case, everything below was read off live production responses on 4 September 2026 — not off a spec, and not off docs that might have drifted. Every field name listed here appeared in a real response.

How to read this

Stocklake exposes 19 MCP tools. Eight are open; eleven require Pro. The split is not "free gets a worse version of everything" — it is closer to market data is open, synthesis is paid. Prices, fundamentals, OHLCV history, the nine standard technical indicators, screening and market movers are all on the free tier at full fidelity. What Pro adds is the derived layer: composite scores, AI-written research, insider and institutional detail, macro and sector assessments, the raw multi-year financial statements underneath the forensic scores, and the second set of nine technical indicators.

Two mechanics are worth knowing before the tables. First, tools/list returns all 19 tools regardless of tier — gating happens when you call, not when you enumerate. An agent that discovers the toolset on a free key will see eleven tools it cannot yet use. Second, a Pro-gated call from a free key does not fail silently: it returns a structured pro_required object carrying a small real preview of the data, plus your_tier, message, upgrade and a tell_your_user string written for an agent to relay verbatim. That is a deliberate design choice — a paywall an agent can explain is more useful than one it can only report as an error.

The tool index

Every tool, its tier, the question it answers, and its hard ceiling. Parameters marked * are required; the rest show their default.

Price, identity and fundamentals

All free, all on get_stock and get_stocks. Forty-four top-level fields, identical for a US mega cap and a European or Japanese listing — there is no reduced schema for international symbols.

earnings_is_estimate is the one to read carefully: when true, earnings_date is a projection, not a confirmed date. And currency reports the currency of the quote you are being given, which for a US-listed depositary receipt of a European company is USD, not the home-market currency — check country and exchange together if that distinction matters to you.

Technical indicators

Eighteen indicators live under indicators on get_stock. Nine are free; nine more unlock on Pro. Each is a nested object, not a bare number — macd carries its line, signal and histogram; bollinger_bands carries upper, middle, lower and width.

A Pro-only signals block sits alongside them, flattening the same readings into labelled verdicts — rsi.signal as "neutral", td_sequential.phase as "countdown_done", vix_fix.percentile, williams_ad.divergence, and an overall roll-up. An indicator with insufficient history returns null rather than being omitted, so a key you expect is always present.

Indicators over time

get_indicator_history (Pro) returns daily snapshots, up to 730 days. Its default response is deliberately slim — seven fields per snapshot — because a 700-row response with fifteen fields each is a lot of context for an agent to carry. Pass full=true to restore the rest.

Note that analyst_rating and analyst_target are in the history payload — which makes this the way to see how analyst posture on a name has shifted, not just where it stands today. If fewer than three snapshots exist for a symbol the response comes back with an empty snapshots array and an explanatory note, rather than an error.

Financial statements

get_stock_financials (Pro) is a separate tool from everything above, not a bigger field on get_stock — it returns the raw balance sheet, income statement and cash flow line items themselves, not a score derived from them. Where forensic_scores tells you a company's Altman Z-Score is 2.1, this tool is where you'd go to see the actual working-capital and total-assets figures that number was built from. period chooses annual (up to 5 fiscal years, the default), quarterly (roughly the last 7-8 quarters), or both; statement narrows to one of the three statements or returns all of them. Around 50-60 curated line items per statement, each shaped {line_item: {period_end: value}} — every available period for a line item comes back in one call, not just the latest one, so you can compute your own year-over-year deltas rather than requesting each year separately. Kept as its own tool deliberately: folding the full annual-plus-quarterly payload into get_stock_research would have roughly tripled that tool's response size for a use case most callers of it don't need.

Price history

get_stock_history (free, up to 365 days) returns bars with date, open, high, low, close, volume — and a source field naming which feed that individual bar came from. Per-bar provenance is unusual to expose, and it exists because bars from different feeds can disagree on volume for the same session.

Scores and synthesis

These are the Pro-only blocks on get_stock — six documented below, plus signals (a flattened view of the same technical readings, covered under indicators above). Together the seven span all ten Pro-only top-level fields (ai_score alone accounts for four: ai_score, ai_score_band, ai_verdict, ai_headline), and they answer genuinely different questions — it is worth not treating them as interchangeable "the AI's opinion".

stance_signals is the one most worth understanding, because it is the only field that shows you disagreement. A single symbol typically returns five rows — technical_rating, ai_summary at two horizons, sentiment_enrichment, analyst_consensus — and they are not required to agree. A name can carry a NEUTRAL swing stance and a POSITIVE long-term stance simultaneously, each with its own as_of timestamp. For an agent, that is more honest input than a single blended verdict would be.

The one-call research bundle

get_stock_research (Pro) is a different shape from everything above: a five-key envelope rather than a flat field list. stock holds a 17-field summary; ai_summary holds verdict, ai_score, ai_score_band, summary, key_points[], risks[], headline, price_at_generation, generated_at, plus near_term and longer_term objects each carrying their own stance and confidence; news and signals hold recent scored items; and sentiment holds signal, signal_score, signal_score_band, insider_trend, institutional_pct. The price_at_generation field matters — it tells you what the price was when the text was written, so you can judge whether the narrative has been overtaken by the tape.

News and sentiment

This is where the tier difference is starkest, and it is worth being blunt about it. Free get_stock_news returns five articles from the last 30 days, and each article has exactly two fields. Pro returns up to 50 from the last 90 days, with six.

The free response carries a tier_note string naming what Pro would add, so an agent can tell a user exactly what it is missing without guessing. Requesting more than the tier ceiling is clamped silently — asking for limit=80, days=400 on a free key returns 5 and 30, and the response echoes those numbers back, so you can always tell what you actually got.

get_news_feed (Pro) is the market-wide counterpart: same six article fields plus symbol, filtered by min_signal_score rather than by ticker — the tool to call when the question is "what is worth reading right now" rather than "what is being said about NVDA".

Insider and institutional

get_insider_activity (Pro). Top level carries signal, signal_score, signal_score_band, insider_signal, inst_signal, summary, insider_buys, insider_sells, inst_ownership, total_holders, holder_count_divergence, updated_at. Beneath that:

The last two fields on a holder are the interesting ones: sources lists which filings agreed on that position, and share_counts_by_source shows the individual numbers. Where they disagree, holder_count_divergence flags it rather than picking a winner. The free-tier preview on this tool still returns real values for signal, insider_signal, inst_signal, insider_buys and insider_sells — enough to know whether the full call is worth making.

Market and sector context

get_market_pulse is free and needs no arguments: vix as a bare number, fear_greed with value / description / last_update, indices with spy / qqq / iwm, bonds_commodities with tlt / gld, and a breadth object giving universe_size, oversold_count, overbought_count, neutral_count and each as a percentage.

get_market_assessment (Pro) returns 26 top-level fields. The regime side gives regime, regime_bias, regime_bias_note, regime_confidence, regime_rationale, regime_strength, regime_stale, regime_updated_at, macro_score, macro_score_trend, risk_appetite_score, vix_at_assessment. The outlook side gives outlook, outlook_conviction, outlook_rationale, outlook_key_risk, outlook_stale, outlook_updated_at, preferred_sectors, avoided_sectors, key_risks, catalyst, watch_for, equity_view. Plus an indicators object holding breadth, macro_data, market_sentiment and volatility_term_structure.

The two _stale booleans are load-bearing: they tell you the assessment has aged past its refresh window, which is a very different situation from it being current and merely unchanged. macro_score_trend gives change_7d, change_30d and a direction label, so you get the trajectory of the read without a second call.

get_sector_intelligence (Pro) returns sector, signal, sector_score, sector_score_trend, strength_score, cycle_stage, rotation_signal, confidence, drivers, alert, updated_at — plus a substantial stats block with avg_rsi, rsi_distribution, rsi_percentile, avg_perf_1w_pct, avg_perf_1m_pct, perf_1m_percentile, cap_weighted_1m_pct, sma200_breadth_pct, top5_concentration_pct, overbought_pct, oversold_pct, median_market_cap_B, top_movers, bottom_movers, stocks, stocks_cleaned, outliers_excluded. Omit sector to get all eleven; add sort_by_strength=true to get them ranked.

Calendars

The earnings pair is the cleanest illustration of the whole tier model: the free tool tells you who reports and when, the Pro tool adds what to watch for. The economic calendar splits its response into released_recent, today and upcoming, so a released event carries a real actual alongside its consensus and previous, and an upcoming one does not.

What a call costs

Quotas are per calendar day, resetting at midnight UTC: guest 25, free 200, Pro 5,000. Guest needs no key at all — drop the Authorization header entirely and you are rate-limited by IP.

The part that surprises people: five tools bill one call per ticker actually returned, not one per request. A get_stocks call for 25 symbols costs 25 against your quota, not 1.

Two consequences worth planning around. get_signals defaults to 25 but will go to 50 if you ask — the default is deliberately lower than the ceiling to keep a typical response light, so you have to opt into the wider feed. And a Pro-gated batch tool called from a free key costs 1, not the batch size: the request is rejected at the gate before it reaches the query, so exploring what you cannot access yet is cheap.

Coverage, and where it thins out

Around 3,500 symbols across US, European and Asian listings. The schema does not change by region — SAP, Toyota and BP each return the same 54 Pro fields as NVIDIA, with populated country, sector, relative_strength, ai_score and stance_signals. What varies is within a field rather than across the field list.

forensic_scores is the honest example. Checked across four symbols, one returned all three formulas, one returned two, and two returned one apiece — the rest coming back as {"score": null, "note": "..."} with a plain-English explanation, most often that the company is a bank or insurer whose balance sheet does not carry the line items those formulas require. That is a real limit of the formulas, not a gap in the data, and the tool says so in the response rather than silently omitting the key. The same pattern holds elsewhere: elliott_wave is frequently null because a clean wave count often is not present, and get_indicator_history returns an empty array with a note when a symbol has fewer than three snapshots.

The honest caveat

This page was read off live responses on one specific day, and the surface it describes moves. Fields get added; defaults get tuned; a cap that reads 25 today was 50 four months ago. Everything here is accurate as of 4 September 2026 and was verified field by field against production, but if you are building something that depends on an exact field being present, call tools/list and read one real response rather than trusting any prose — including this. Nothing on this page is a recommendation about any security; it is a description of what an MCP tool returns.

Frequently asked questions