What only Genka has

The Ecosystem page tells you what other services to use when Genka isn't the right fit. This page flips that — the five primitives that aren't in any free Indian-markets API today, with the curl that proves it.

1. Concall corpus depth + BM25 ranking

17,979 transcripts · 483 NSE 500 symbols · 8 years (2018-08 → 2026-04). Every search hit returns kw_hits + tf_density + days_old + a highlighted snippet so you can see why it ranked. kind=IP filter narrows to slide-deck-only hits where order-book numbers actually live.

curl -sH "$H" "https://genka.dev/latest/concalls/search?q=margin+expansion&limit=5"
# => SYRMA (9.1) -> BATAINDIA (5.9) -> BIOCON (5.4) -> IGIL (5.3) -> INDHOTEL (5.1)

Where else? AlphaSense + Bloomberg have it for 100× the price. Free Indian alternatives (Screener, Tickertape) don't index transcripts.

2. FII / DII F&O participant breakdown

Daily flow split across six segments: FII Cash, DII Cash, FII Index Futures, FII Index Options, FII Stock Futures, FII Stock Options. Plus 5d / 30d rolling and a 30-day cumulative time-series with Nifty + Sensex close per row for chart-against-flow analysis.

curl -sH "$H" "https://genka.dev/latest/fiidii/latest" | jq '.data | {fii, dii, rolling}'
# => fii_opt_30d: -83,053 cr  <-- the actionable signal

Where else? Most free APIs give the headline cash number. The F&O participant split exists in NSE's PDF reports daily but no other free JSON API parses it.

3. Holdings reverse-index `/companies/{X}/holders`

Forward direction (which stocks does fund X hold?) is everywhere. Reverse direction — which funds hold stock X, ranked by weight — isn't. Returns each fund + weight + as_of, plus /holders/summary for total INR institutional exposure.

curl -sH "$H" "https://genka.dev/latest/companies/JINDALSTEL/holders" | jq '.data | sort_by(-.weight_pct) | .[:5]'
# => the 5 highest-conviction MF holders of JINDALSTEL today

Why it's hard. Requires (a) parsed monthly factsheets across all funds, (b) ticker resolution from instrument-name strings, (c) a lookup index on the (ticker, fund) pair. We do all three.

4. AMC factsheet PDF history via Gemma-27B parse

Tickertape covers ~2,094 funds with one current snapshot. PPFAS publishes monthly factsheet PDFs back to 2014. We parse the PDFs through OpenRouter Gemma-27B for OCR + structured extraction; result: 124 monthly snapshots for PPFAS Flexi Cap alone, plus HDFC + ICICI factsheet history coming online via the multi-source ingest. WayBack Machine fallback for older months Tickertape never crawled.

curl -sH "$H" "https://genka.dev/latest/mf/ppfas-parag-parikh-flexi-cap/holdings?as_of=2022-09-30"
# => time-travel to a specific monthly snapshot, not just "today"

Source rank. amc_pdf_* > wayback > tickertape > factsheet > mfdata_in. Reads dedup via the multi-source view; provenance is on every row.

5. 23-tool MCP server, spec 2025-06-18 compliant

The same data via official @modelcontextprotocol/sdk Streamable HTTP transport. Tracks the MCP spec by upstream — we don't hand-roll the JSON-RPC layer. 23 tools across search / capabilities / mf_* (8) / co_* (4) / prices_* (3) / concall_* (3) / sector_get. Every tool is a thin loopback into the REST handler so auth + billing + audit run through one pipeline.

claude mcp add --transport http genka https://genka.dev/mcp \
  --header "x-api-key: $GENKA_API_KEY" --scope user
# Claude Code: "find me momentum funds with sharpe > 1, then pull holdings of the top 3"

Where else? Most public APIs ship REST only. MCP is for the agent that wants typed tools, not URL-string composition.

What we don't have