Step-by-step of getting live NSE option-chain data via MCP — the exact setup I used for the NSE research skill. No API key, no auth circus. Agar tum NIFTY/BANKNIFTY data chahte ho bina paid vendor ke, ye read karo.
What nse-bse-mcp gives
A local MCP server (HTTP :3000) exposing 57 NSE tools: `nse_fno_historical`, `nse_option_chain`, `nse_filtered_option_chain`, `nse_compile_option_chain`, `nse_vix_historical`, `nse_download_fno_bhavcopy`, `nse_get_market_status`. Ye sab NSE ke public endpoints wrap karte hain. No API key required.
Install (verified on macOS, node 22)
npx -y nse-bse-mcp
Server `http://localhost:3000/mcp` pe start hota hai. Background mei run karo taaki terminal band hone par na mare:
nohup npx -y nse-bse-mcp > /tmp/nse-mcp.log 2>&1 &
Health check:
curl -s -X POST http://localhost:3000/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
57 tools list aayenge. Agar empty aaye toh server nahi chala — log check karo.
Fetch live chain (Python)
import urllib.request, json
def fetch_snapshot(symbol="NIFTY"):
payload = {"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"nse_filtered_option_chain",
"arguments":{"symbol":symbol,"strike_range":5,"max_items":50}}}
req = urllib.request.Request(
"http://localhost:3000/mcp",
data=json.dumps(payload).encode(),
headers={"Content-Type":"application/json",
"Accept":"application/json, text/event-stream"})
resp = urllib.request.urlopen(req, timeout=20).read().decode()
if "data:" in resp:
resp = resp.split("data:")[-1]
data = json.loads(resp)
text = data["result"]["content"][0]["text"]
s = text.strip()
j = json.loads(s[s.find('{'):s.rfind('}')+1])
chain = j.get("data") or j.get("_metadata",{}).get("data") or []
rows = []
for strike in chain:
ce = strike.get("CE",{}); pe = strike.get("PE",{})
rows.append((strike["strikePrice"], "CE", ce.get("openInterest"),
ce.get("impliedVolatility"), ce.get("lastPrice")))
rows.append((strike["strikePrice"], "PE", pe.get("openInterest"),
pe.get("impliedVolatility"), pe.get("lastPrice")))
return rows
Real result (19-Aug-2026 09:31 IST)
NIFTY 50 = 24086.25 (Open, -0.28%)
42 rows (21 strikes × CE+PE)
Top OI strike: 24500 CE
Sample CE OI: 3761 | CE IV: 12.36% | CE last: 142.50
Ye numbers NSE live feed se aaye hain. Mock nahi.
Gotchas (real bugs I hit)
1. **BANKNIFTY `range:10` error** — "Response is too large" deta tha kyunki BANKNIFTY mei zyada strikes hain. Fix: `strike_range` + `max_items` use karo (NIFTY mei `range:10` chal gaya, BANKNIFTY mei nahi).
2. **BANKNIFTY response `_metadata` wrapper mei** — NIFTY clean JSON deta hai, BANKNIFTY metadata mei. Parser dono handle kare (upar wala code karta hai).
3. **Expiry format** — `"25-Aug-2026"` JSON date nahi, parse fail. Normalize `2026-08-25` karo.
4. **CE/PE nested** — structure `{data:[{strikePrice, CE:{...}, PE:{...}}]}`. Direct `json.loads(text)` fail hoga, iterate karo.
Store to DuckDB
import duckdb
con = duckdb.connect("options.duckdb")
con.execute("""CREATE TABLE IF NOT EXISTS market_raw (
session_date DATE, symbol VARCHAR, strike DOUBLE, type VARCHAR,
open_interest BIGINT, iv DOUBLE, last_price DOUBLE, spot DOUBLE)""")
rows = fetch_snapshot("NIFTY")
spot = 24086.25
con.executemany("INSERT VALUES (?,?,?,?,?,?,?,?)",
[(date.today(), "NIFTY", r[0], r[1], r[2], r[3], r[4], spot) for r in rows])
Comparison: free vs paid
| Source | Cost | Latency | Setup |
|--|||-|
| nse-bse-mcp | ₹0 | ~1-2s | npx command |
| NSE official API | ₹0 but unstable | varies | auth circus |
| Paid (TrueData etc) | ₹1500+/mo | <1s | subscription |
| Manual screenshot | ₹0 | manual | time sink |
Free MCP enough hai research/paper-mode ke liye.
FAQ
**Q: API key chahiye?** Nahi. NSE public data hai.
**Q: Windows pe chalega?** Haan, npx cross-platform hai.
**Q: Kitni baar call kar sakte hain?** NSE rate-limit hai, 5-min interval safe hai.
**Q: BANKNIFTY bhi?** Haan, `strike_range:5, max_items:50` se 102 rows aate hain.
Common mistakes
1. **Background na chalaana** — terminal band, server mara, "MCP down" error.
2. **BANKNIFTY = NIFTY treat karna** — alag response format hai.
3. **Expiry format assume ISO** — crash karega.
4. **Har second call** — rate-limit ban.
What I learned
Data pipeline banane mei 80% time debugging format issues mei gaya, na ki logic mei. NSE ka response consistent nahi hai across symbols — defensive parser likhna padta hai.
My workflow
*Data via NSE public site. Research only, not live-trading.*
Deep dive: response parsing why it's tricky
NSE MCP ek consistent API nahi deta. NIFTY ke liye `nse_filtered_option_chain` direct JSON deta hai jisme `data` array hai. BANKNIFTY ke liye same call `_metadata` wrapper deta hai jisme actual data `_metadata.data` mei hai, aur upar ek message hota hai "Response is too large — use filters." Matlb BANKNIFTY mei zyada strikes hain, isliye server ne truncate kar diya aur metadata mei wrap kar diya.
Isliye mera parser dono check karta hai:
chain = j.get("data") or j.get("_metadata",{}).get("data") or []
Agar `data` nahi mila toh `_metadata.data` try karta hai. Fir CE/PE nested objects iterate karke rows banata hai.
Expiry handling
Option chain mei multiple expiries hoti hain. MCP default nearest expiry deta hai. Agar specific expiry chahiye:
arguments={"symbol":"NIFTY","expiryDate":"2026-08-25","strike_range":5}
Dhyan rahe: format `25-Aug-2026` FAIL karega (string date JSON nahi). Use `2026-08-25` (ISO). Maine normalize function banaya hai jo `DD-Mon-YYYY` ko `YYYY-MM-DD` mei convert karta hai.
Error handling in production
try:
rows = fetch_snapshot("NIFTY")
except urllib.error.URLError as e:
print("MCP down, retrying in 60s")
time.sleep(60)
# ya launch script se restart karo
Mera `launch_recorder.py` har run pe check karta hai ki server up hai; nahi toh `npx -y nse-bse-mcp` start kar deta hai. Isse Cron job 24x7 chalta hai bina manual intervention.
Rate limits aur etiquette
NSE public endpoints rate-limit karte hain. 5-minute interval pe 1 call safe hai. Agar `429` ya empty response aaye, backoff karo (exponential: 60s, 120s, 240s). Mass scraping mat karo — IP ban ho sakta hai.
Security
Server `localhost` pe hi rakho. `0.0.0.0` pe expose mat karo — koi bhi network se tumhara NSE data pull kar sakta hai. Firewall default block hai, usko mat kholo.
Real use case: my NSE research
Is setup se main daily 42 (NIFTY) + 102 (BANKNIFTY) rows collect karta hoon DuckDB mei. 4 weeks mei 20+ sessions build hote hain, tab signal gate real probability nikalta hai. Bina is data ke koi model blind hai.
*Data via NSE public site. Research only, not live-trading.*
Troubleshooting checklist
| Symptom | Cause | Fix |
||-|--|
| Empty tools/list | Server nahi chala | `nohup npx -y nse-bse-mcp &` |
| "too large" on BANKNIFTY | Range wide | `strike_range:5, max_items:50` |
| `_metadata` in response | BANKNIFTY truncate | Parser handles it |
| Expiry parse error | `25-Aug-2026` format | Use `2026-08-25` |
| 429 empty | Rate limit | Backoff 60s+ |
| CE/PE missing | Nested structure | Iterate CE{} PE{} |
Comparison with Indian-Option-MCP
Doosra option: `npx -y indian-option-mcp` (stdio mode). Difference:
Main dono use karta hoon — nse-bse-mcp primary (HTTP stable), indian-option-mcp backup.
What this unlocks
Live NSE data milne se:
1. Real OI/IV features bante hain (no fake data).
2. Max pain, PCR, IV skew calculate kar sakte ho.
3. Signal gate ko real feed milta hai.
4. Backtest realistic ban jata hai.
Bina data ke ML sirf theory hai. Ye setup theory ko practice banata hai.
*Data via NSE public site. Research only, not live-trading.*
My honest setup notes
Maine ye setup 19-Aug-2026 ko banaya tha. Pehle 2 ghante sirf format bugs fix karne mei gaye — NIFTY clean tha, BANKNIFTY ne sab bigada. Lesson: **hamesha dono symbols test karo** ek hi function mei, mat assume karo ki ek kaam karega toh doosra bhi karega.
Ab ye Cron job pe hai: `*/5 9-15 * * 1-5` — market hours mei har 5 minute. Monday se Friday. Saturday/Sunday band (NSE closed). EOD snapshot 16:00 pe alag job karta hai.
Data DuckDB mei accumulate ho raha hai. Abhi 1 session tha, abhi NO_TRADE gate sahi reject kar raha hai. 4 weeks baad 20+ sessions honge, tab walk-forward validator pehli real signal degi.
Final recommendation
Agar tum NIFTY/BANKNIFTY par research kar rahe ho aur paid data afford nahi kar sakte, nse-bse-mcp best free option hai. 30 minute mei setup ho jata hai agar upar ke gotchas yaad rakho. Paid vendors tab consider karo jab live sub-second execution chahiye.
*Data via NSE public site. Research only, not live-trading. SEBI compliance separate topic.*
Quick start (copy-paste)
# 1. Start server
nohup npx -y nse-bse-mcp > /tmp/nse-mcp.log 2>&1 &
# 2. Verify
curl -s -X POST http://localhost:3000/mcp -H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head
# 3. Run fetch (use code above)
python3 fetch_nifty.py
Bas. 10 minute mei live NSE data tumhare paas hai. Baaki articles mei hum isi data pe features aur signals banayenge.
*Data via NSE public site. Research only, not live-trading.*
Disclaimer
Sab kuch research purpose ke liye hai. Live trading ke liye SEBI registration aur risk management alag se chahiye. NSE data public source se aata hai, accuracy guaranteed nahi.
Agar setup mei atak gaye, comment karo — main exact error dekh kar fix bata dunga.