Choose endpoints

Resolve a name with /v2/search, confirm a symbol with /v2/coverage, then call only the family the question needs. Most questions start at /v2/snapshot. This page is the table you check before writing a workflow, an agent policy, or a batch job.

​

The table

Find the user’s question, take the route plan, stop there. Every MCP tool is named stockcontext_<operation> and takes the same arguments as native JSON (lists are arrays, not CSV), so GET /v2/compare?symbols=NVDA,AMD becomes stockcontext_compare({ "symbols": ["NVDA", "AMD"] }). The mapping is mechanical; the column below names the tool where it isn’t obvious. Starter+ means a paid plan: any route past the Free four is callable on both Starter and Builder. Free keys can call /v2/search, /v2/coverage, /v2/snapshot, and /v2/valuation; everything else returns 403 PLAN_UPGRADE_REQUIRED on a Free key. The route list, rate limits, and quotas live on Plans and limits; check there, not here.

​

Snapshot or valuation?

These two answer different questions, and the difference decides whether you make one call or two. /v2/snapshot is breadth: a quote, multi-window performance, the 52-week range, and a fundamentals_quick block. The four headline multiples there — pe_ttm, ps_ttm, pb, ev_ebitda_ttm — each ship as a {value, vs_5y} object (the number is at .value; vs_5y is a near_5y_low … near_5y_high band), while fcf_yield_pct, dividend_yield_pct, and the ttm_* / shares_outstanding / beta_5y_weekly fields stay flat floats. One call answers “what’s going on with this stock?” including a first read on whether each multiple sits high or low in its own range. /v2/valuation is depth on those multiples: each one arrives as {value, vs_own_history}, where vs_own_history carries percentiles across 3y/5y/10y windows, a magnitude anchor median, and a window-scoped label (near_5y_low through near_5y_high). P/E, P/S, and the anchor also carry a change block that decomposes the move into price versus fundamentals. That’s how you say “low or high versus its own history, and why” — which a bare pe_ttm of 33 cannot tell you. The label ranks the multiple against the symbol’s own past only; it is never a cheap/expensive verdict versus peers or fair value. Add the second call when the user asks whether the price is justified, not only what it is. For “is AAPL up today?” snapshot is the whole answer. For “is AAPL expensive?” you want both, and the valuation check guide walks the two-call path on a Free key.

​

Before you store or batch a symbol

Call /v2/coverage?symbol= once for user-entered or stored tickers. It tells you the symbol is in the 5,628-symbol universe and which data families it supports: foreign private issuers return unsupported for /v2/filings and /v2/insider, and ETFs skip stock-only families. The full universe and its gaps are documented in Coverage and gaps; skip the check and you find the gap later, mid-answer.

​

Where StockContext’s data ends

StockContext is focused on US-equity context: quotes, performance, valuation, SEC filings, SEC events, insider Form 4/Form 144 activity, and the fixed ETF tier’s N-PORT holdings where collected. Segments, ownership/13F/13D-G, and governance are private beta rather than default public-core inputs. It does not return options or implied vol, analyst price targets or consensus estimates, real-time tick data, news, transcripts, non-US local listings, or execution, so an agent always knows exactly where the data ends and never guesses. If a prompt needs one of those, no route here answers it: say so rather than blending in a guess. Two habits waste quota and degrade answers. Calling every family “just in case” makes an agent more likely to stitch a blended claim from fields it didn’t need. And calling /v2/fundamentals for a valuation question, or /v2/history for a timing question, reaches past the route that already answers it: /v2/valuation and /v2/technicals respectively.