eCourts API: Search 35 Crore+ Indian Court Cases by CNR, Party or Advocate

Last updated: 3 November 2026

Key takeaways

• Coverage: the Supreme Court, all 25 High Courts, district and taluka courts in 36 states and UTs, and 18 tribunal and commission families. A search with no filters returned 35,23,12,960 records on 3 November 2026.
• Use Case Search v2 at GET /api/partner/v2/search. It takes the same parameters as v1, adds a facetMode field, and costs the same. If you are on v1, move off it now.
• Five groups, 32 endpoints: Cases 12, Cause lists 4, Electoral 3, LegalCheck 6, Acts and Sections 7. Court structure, enums and the facet catalog are free.
• Real costs: 1 lakh searches cost ₹60,000 at the Pay As You Go rate or ₹20,000 at the Enterprise base rate, before GST. Failed requests and empty results are not billed.
• Data, not judgment: a name match is not identity, pending is not guilt and disposed is a status, not an outcome. Read the order before you rely on a case.

What is the “eCourts API”, and how is eCourtsIndia different from eCourts?

“eCourts” is the Government of India’s eCourts project, whose eCourts Services portal is run by the e-Committee of the Supreme Court and hosted by NIC; the portal offers no developer API or API keys. eCourtsIndia is a separate, private platform built by Billion Insights that indexes public court records and sells API access to them.

People search for “eCourts API” because the official portal is where case status lives. The portal is free and authoritative. What it does not offer is a way for software to ask questions across courts: every search is one state, one district, one court complex and one captcha at a time. Several companies sell an “eCourts API” for that reason, including Surepass, whose eCourts API page lists inputs such as the name of the individual or entity, father’s name, court, address and state, and offers Get API Key and Contact Sales buttons with no price shown (read 3 November 2026).

eCourtsIndia describes itself in its own footer as “A non Govt. Initiative to use AI in Law”. It is not affiliated with any court or with the e-Committee. What it adds is one index across every tier, AI summaries of orders, cause lists, a statute book and an identity-matched background check, all behind one key. For the bigger picture of who holds what, see our map of India’s court data stack.

What does the API cover?

As of 3 November 2026, the index held 35,23,12,960 case records: 11,98,705 from the Supreme Court, 5,94,09,054 from High Courts, 28,63,90,533 from district and taluka courts and 53,12,844 from tribunals. The homepage also lists 141.1 crore orders.

Tier	Records	What is included
Supreme Court	11,98,705	Supreme Court of India (court code SCIN01)
High Courts	5,94,09,054	All 25 High Courts and their benches (for example DLHC01, HCBM01, HCMA01)
District and taluka courts	28,63,90,533	Civil, criminal and family courts in 36 states and union territories
Tribunals and commissions	53,12,844	18 families: NCLT, NCLAT, ITAT, CESTAT, DRT, DRAT, NGT, AFT, SAT, TDSAT, APTEL, CAT, CCI, GSTAT, RCT, consumer commissions and others
All records	35,23,12,960	Evergreen figure used on this blog: 35 crore+

Two coverage facts trip up new users. Cause lists cover district and taluka courts only (8,000+ courts, about 30 days of lists); High Court and tribunal cause lists are not in the dataset. And tribunal records are searched by court code, because there is no “court type” filter. Our tribunal guides for NCLT and NCLAT and DRT and DRAT list the codes.

What are the 32 endpoints?

The docs list 32 endpoints in five groups: Cases (12), Cause lists (4), Electoral (3), LegalCheck (6) and Acts and Sections (7).

Group	Endpoints	Main paths	Price
Cases (12)	Case Detail, Case Search, Case Search v2, Facet Catalog Full and Search, Order Download, Order + AI, Order Markdown, Case Refresh, Bulk Refresh, Bulk Refresh Status, Enum Reference	GET /api/partner/v2/search; GET /api/partner/case/{cnr}; GET /api/partner/case/{cnr}/order-ai/{filename}; GET /api/partner/case/{cnr}/order-md/{filename}; POST /api/partner/case/{cnr}/refresh; POST /api/partner/case/bulk-refresh	Search ₹0.60 / ₹0.20; detail ₹1.50 / ₹0.50; order PDF ₹3.75 / ₹1.25; order + AI ₹7.50 / ₹2.50; markdown ₹5.25 / ₹1.75; refresh ₹0.15 / ₹0.05 per CNR
Cause lists (4)	Court Structure, Cause List Search, CNR Check Batch, Available Dates	GET /api/partner/causelist/search; POST /api/partner/causelist/cnr/batch	Search ₹3.00 / ₹1.00; batch ₹0.30 / ₹0.10 per CNR; structure and dates free
Electoral (3)	Roll Search, EPIC Lookup, Capabilities	GET /api/partner/electoral/search; GET /api/partner/electoral/epic/{epic}	Not on the public rate card
LegalCheck (6)	Submit, Status, Report, Search Log, List, Models	POST /api/partner/legal-check; GET /api/partner/legal-check/{code} and /report	₹99 / ₹33 per accepted job; reads free
Acts and Sections (7)	Case Provisions, Get a Section, Sections Batch, Section Equivalents, Resolve Act Name, Search Sections, Corpus Status	GET /api/partner/case/{cnr}/provisions; GET /acts/{key}/sections/{n}	₹0.15 / ₹0.05 per provision; corpus status free

How do you authenticate and make your first call?

Send your token in an Authorization header as “Bearer eci_live_…” to https://webapi.ecourtsindia.com; tokens are 41 characters including the eci_live_ prefix. Sign up at ecourtsindia.com/api, copy the token from the dashboard and keep it in an environment variable.

Every response comes in a data envelope with a meta block carrying a request_id. Log it; support asks for it. A call with a missing or wrong token fails cleanly. When we called the partner API without a valid token on 3 November 2026, it returned HTTP 401 with this body:

{"error": {"code": "INVALID_TOKEN", "...": "..."}, "meta": {"request_id": "..."}}

How does Case Search v2 work?

GET /api/partner/v2/search takes the same parameters as v1 and returns the same shape, plus data.facetMode, which tells you whether facet counts were computed. Price and the 200-row page limit are unchanged.

The one real difference is facets. v2 counts facets only when a search matches under 10 lakh cases, or under 1 lakh for a relevance-sorted text or name search. When it skips them, facetMode is “full” and facets is an empty object; when the skip is because a relevance-sorted text or name search passed 1 lakh matches, totalHits is a lower bound of 100000 rather than an exact count. When facets are counted, facetMode is “counted”. v2 also ignores maxFacetValues and returns up to 200 values per facet. The docs mark v1 as “Deprecated soon” with no shutdown date; migration is a URL change plus a facetMode check.

Parameter	What it does	Example
query	Full-text search over case metadata and the text of orders; supports “phrases”, AND / OR / NOT and a trailing *	“Negotiable Instruments” AND 138
litigants	Party name on either side	Kingfisher Airlines
petitioners, respondents	One side only	State Bank of India
advocates, judges	Advocate or judge name	Sharma
nameMatchMode	all (default), any, phrase or fuzzy, for the name fields	phrase
courtCodes	Exact court codes; tribunals are searched this way	DLHC01, NCLTMB0
courtLevels	Tier filter	SC, HC, DC, TRIBUNAL
stateCodes, districtCodes	Two-letter state codes and numeric district codes	UP and 24 (Lucknow)
caseTypes, caseStatuses	Code values, not English names	CC, WP_C; PENDING, DISPOSED
caseNumbers	Exact match on registration or filing number	497/2015
filingDateFrom, filingDateTo (and other date pairs)	Date ranges in YYYY-MM-DD	2024-01-01
actKeys, provisions, provisionPaths	Statute filters listed in the v2 docs	Use the facet catalog for values
facets, sortBy, page, pageSize, fields	Drill-down counts, sorting, paging (pageSize up to 200) and projection	facets=caseStatus

Three mistakes return zero rows silently. A High Court code without its bench suffix (DLHC instead of DLHC01). An NCLT code without its trailing zero (NCLTMB instead of NCLTMB0). And an English case type (“CIVIL”) instead of a code. Use the free Facet Catalog endpoints, GET /api/partner/v2/facets/full and GET /api/partner/v2/facets/search, to turn a name into an exact filter value. For multiple values on REST, repeat the key: courtCodes=DLHC01&courtCodes=HCBM01. The case type encyclopedia explains what the codes mean.

import os
import requests

BASE = "https://webapi.ecourtsindia.com"
HEADERS = {"Authorization": f"Bearer {os.environ['ECI_TOKEN']}"}

params = [
    ("litigants", "HDFC Bank"),
    ("query", '"Negotiable Instruments" AND 138'),
    ("stateCodes", "UP"),
    ("districtCodes", "24"),        # Lucknow
    ("facets", "caseStatus"),
    ("pageSize", "50"),
]
r = requests.get(f"{BASE}/api/partner/v2/search", headers=HEADERS,
                 params=params, timeout=60)
r.raise_for_status()
data = r.json()["data"]
print(data["totalHits"], data.get("facetMode"))
for row in data["results"]:
    # results carry no title: build one from the party arrays
    title = f"{', '.join(row.get('petitioners', []))} vs {', '.join(row.get('respondents', []))}"
    print(row["cnr"], row.get("caseType"), row.get("caseStatus"), title)

Worked example: a bank’s cheque-bounce complaints in one district

We ran this search through the eCourts MCP, which queries the same index, on 3 November 2026: litigant “HDFC Bank”, full text “Negotiable Instruments” AND 138, state UP, district 24 (Lucknow). It returned 15,443 records. The facets showed 7,901 pending and 7,542 disposed; 15,304 of the records were criminal complaint cases (case type CC); and filings peaked in 2017 with 4,294. The first page came from the Chief Judicial Magistrate courts in Lucknow (court code UPLK04), including the dedicated N.I. Act courts.

The litigants filter matters here: the bank is the complainant in these cases, so “15,443 records naming HDFC Bank” says nothing bad about the bank. Direction and role always come first. Our Section 138 tracking guide builds a full tracker on this pattern.

Worked example: a tribunal by court code

A search with only courtCodes=NCLTAL0, the NCLT bench at Allahabad, returned 6,863 records. Interlocutory applications under the IBC led with 3,435, followed by 1,042 company petitions under the IBC and 587 company petitions under the Companies Act; 4,517 were disposed and 2,346 pending. A second search, litigant “State Bank of India” as a phrase with courtLevels=TRIBUNAL, returned 1,56,364 records, and the six largest court codes in the facet were all DRT benches.

How do you read a case and its orders?

GET /api/partner/case/{cnr} returns the full record: parties, advocates, judges, hearing history, orders, interlocutory applications and tagged matters. The three order endpoints then return the PDF, the text as Markdown, or the text plus an AI analysis.

CNR is the 16-character case number every Indian court record carries, four letters and twelve digits; CNR Number Decoded explains each part. We fetched DLHC010813232015, a Delhi High Court suit by Veiling Ltd and others against Kingfisher Airlines Ltd and others. Trimmed:

{
  "data": {
    "courtCaseData": {
      "cnr": "DLHC010813232015",
      "caseTypeRaw": "CS(OS)",
      "registrationNumber": "497/2015",
      "caseStatus": "DISPOSED",
      "filingDate": "2015-02-23",
      "decisionDate": "2018-11-30",
      "caseDurationDays": 1376,
      "petitioners": ["Veiling LTD", "ORS"],
      "respondents": ["Kingfisher Airlines LTD", "ORS"],
      "orderCount": 16,
      "judgmentCount": 1,
      "hearingCount": 12,
      "iaCount": 9,
      "judgmentOrders": [
        {"orderDate": "2018-11-30", "orderType": "Final Order", "orderUrl": "order-16.pdf"}
      ]
    }
  },
  "meta": {"request_id": "..."}
}

We then called the Order + AI endpoint on order-16.pdf. The response opened with a date check (“the order index and the document text agree this is the 2018-11-30 order”), then gave the full extracted text and a structured analysis: outcome Allowed (Partially - Suit Decreed with Compromise), provisions cited (Order XXIII Rules 1 and 3 and Section 151 of the CPC), and directions to the registry.

Read the text, not just the summary. The order says the suit is “dismissed as withdrawn qua defendant nos. 1 to 3”, decreed in terms of two joint compromise applications signed with defendant nos. 4 to 8 and defendant no. 9, and that defendant nos. 10 and 11 were deleted at the plaintiffs’ request. Kingfisher Airlines is the first-named defendant in the cause title. The AI’s plain-language summary said that “Veiling Ltd and Kingfisher Airlines Ltd reached a settlement” and that Veiling Ltd “won its case against most of the defendants”. The order text supports neither line: the compromise applications were signed by the plaintiffs and defendants 4 to 9, not by Kingfisher Airlines, and a decree in terms of a compromise records what the parties agreed, not who won. The response carries the full text for exactly this reason. Order + AI takes 10 to 60 seconds the first time and is cached after that; Order Markdown can take up to 300 seconds and may return 429 TOO_MANY_CONVERSIONS, in which case wait for the Retry-After time.

How does refresh work, and when should you use it?

POST /api/partner/case/{cnr}/refresh queues a fresh read from the court and returns 202 with status QUEUED; the docs say to fetch the case again after 5 to 10 seconds. Bulk Refresh takes 2 to 50 CNRs and Bulk Refresh Status (1 to 50, free) tells you when each has landed.

Refresh when a pending case shows a next hearing date in the past. That is the signal that the record may have moved. We tested it on a pending 2025 execution petition against The Kingfisher Airlines Ltd at the Patiala House Courts, New Delhi, whose record showed a next date of 7 February 2026. The refresh landed in 21 seconds. Nothing changed: the court’s own record had not published a newer date. That is a useful result in itself. A refresh fixes our copy; it cannot make the court publish faster.

Every bulk refresh request is charged, so poll the status endpoint before asking for the same CNR again. Failed refreshes do not cost you. The refresh lifecycle is covered in detail in from scraping to production.

How do you check cause lists by CNR?

POST /api/partner/causelist/cnr/batch takes 1 to 100 CNRs in a cnrs array and tells you, for each, whether it appears on an available cause list and when. It bills per distinct CNR and ignores duplicates and blanks; over 100 returns 400 BATCH_SIZE_EXCEEDED.

We sent three CNRs on 3 November 2026. A bank’s cheque-bounce complaint in Lucknow was listed for 30 November 2026 in Additional Court No. 2 (N.I. Act). A second Lucknow complaint came back not listed, and because we hold that court’s lists, that means nothing is scheduled in the window. The Delhi High Court suit came back as “no cause-list data”, which means we do not hold High Court lists, not that the case has no hearing. Your code must treat those two answers differently. Cost: 3 × ₹0.30 = ₹0.90 on Pay As You Go.

For a search across lists, GET /api/partner/causelist/search takes q (case numbers, parties, advocates), a date or date range, and filters such as courtComplexCode, court, courtNo, litigant and listType. It needs at least one filter, returns up to 200 rows (default 100) and pages by offset. Our guide to case alerts shows the same data as WhatsApp and email alerts.

What do the Acts and Sections endpoints do?

They return the text of Indian statutes, 4,20,399 provisions in all, with offence classification and old and new code equivalents (IPC to BNS, CrPC to BNSS, Evidence Act to BSA), at ₹0.15 per provision on Pay As You Go.

We read Section 138 of the Negotiable Instruments Act, 1881. The text came back with its footnotes: a dishonoured cheque is an offence punishable “with imprisonment for a term which may be extended to two years, or with fine which may extend to twice the amount of the cheque, or with both”, provided the cheque was presented within six months, a written demand was sent within thirty days of the bank’s memo, and the drawer did not pay within fifteen days of the notice.

The Case Provisions endpoint, GET /api/partner/case/{cnr}/provisions, attaches that text to the citations a case record carries. On the Lucknow complaint it returned an empty list and no charge, even though the search record shows “Negotiable Instruments Act | Section 138”. Parsed citations are not yet filled in for most cases, so fall back to the actsAndSections field and Get a Section. Our IndiaCode JSON API guide and the IPC to BNS mapping go further.

What about the LegalCheck and electoral roll endpoints?

LegalCheck runs an identity-matched litigation background check on a person or company; the electoral endpoints search published voter rolls. Both sit on the same key.

LegalCheck is a job: POST /api/partner/legal-check with an Idempotency-Key header, then poll GET /api/partner/legal-check/{code}, honouring Retry-After: 5, and read /report when it is done (409 NOT_READY before that). The docs give an estimate of about 90 seconds and say it is advisory, not an SLA. A partner can have three jobs queued or running at once. It costs ₹99 on Pay As You Go or ₹33 with a subscription, charged once when the job is accepted; reads are free. The LegalCheck API guide documents the report schema.

The electoral group has Roll Search, EPIC Lookup and Capabilities. Paging runs 1 to 100 for page and pageSize, with at most 9,900 records per query, and a search with no primary criterion returns 400 MISSING_SEARCH_CRITERIA. These records are about real voters, so we show no examples here; use them only for a lawful purpose you can document.

How much does the eCourtsIndia API cost?

Pay As You Go prices are three times the base rate that Enterprise plans pay: a case search is ₹0.60 or ₹0.20, a case detail ₹1.50 or ₹0.50, before 18% GST. Signup gives ₹200 of credit with no card.

Pay As You Go uses prepaid packs of ₹1,000, ₹2,500 or ₹5,000 (plus GST) and the credits never expire. Enterprise Monthly costs ₹10,000 a month plus GST for 10,000 credits a month at the base rate; Enterprise Annual costs ₹1,00,000 a year plus GST for the same 10,000 credits a month. Unused Enterprise credit does not roll over. Failed requests and empty results are not charged on most endpoints; Search Sections is charged per call even with no hits.

Searches a month	Pay As You Go at ₹0.60	Enterprise base rate at ₹0.20
1,000	₹600	₹200
10,000	₹6,000	₹2,000
1,00,000	₹60,000	₹20,000
10,00,000	₹6,00,000	₹2,00,000

Enterprise Monthly is ₹10,000 a month plus GST with 10,000 credits; ask sales for volume beyond the plan. Real workloads mix calls, so price your own mix. A monitoring job that refreshes 1,000 CNRs once a week and reads each case after refresh uses about 4,000 refreshes and 4,000 case details a month: ₹600 plus ₹6,000 on Pay As You Go, or ₹200 plus ₹2,000 at base. Annual plans and custom volumes go through hi@ecourtsindia.com.

What errors and timings should you plan for?

Plan for 401 on a bad token, 400 on bad input, 409 on a LegalCheck report that is not ready, 429 when the order converter is busy, and multi-second waits on anything that touches a court server.

Code or call	Meaning	What to do
401 INVALID_TOKEN	Missing or malformed Bearer token	Check the header format and the 41-character token
PAGE_SIZE_EXCEEDED	pageSize above 200 on search	Lower it; not billed
400 INVALID_CNR	Cause-list batch with no usable CNR	Validate CNRs before sending
400 BATCH_SIZE_EXCEEDED	Cause-list batch above 100 CNRs	Split the batch
400 MISSING_SEARCH_CRITERIA	Electoral search with no primary field	Add a name, EPIC or other criterion
409 NOT_READY	LegalCheck report still running	Poll status, honour Retry-After
IDEMPOTENCY_CONFLICT	Same Idempotency-Key sent with a different body	Use a new key for a new job
429 TOO_MANY_CONVERSIONS	Order Markdown converter is saturated	Wait for the Retry-After duration
503 on facet catalog	Catalog not ready yet	Retry; it is cached for one hour once built
Case refresh	Queued with 202	Re-read after 5 to 10 seconds (ours landed in 21)
Order + AI	First fetch reads the court PDF	10 to 60 seconds, then cached
Order Markdown	PDF to text conversion	Up to 300 seconds

Retry server errors with backoff and jitter; do not retry a 400, fix it. Never put the token in client-side code.

Can you use the API from Claude, ChatGPT or Cursor?

Yes. Add https://mcp.ecourtsindia.com/mcp as a custom connector in your AI client and click Connect to sign in with eCourtsIndia; the docs say not to run a local stdio server. The MCP uses your existing API credits, so the same prices apply.

We ran every example on this page through the MCP. Our setup guides for Claude and for Claude, ChatGPT and Cursor across all three Billion Insights servers have the steps.

What does the API not tell you?

It tells you what the court record says, as of when it was read. It does not tell you who a person is, what a case means for them, or whether the court’s own record is complete.

• A name match is not identity. An exact-phrase search for “Rahul Sharma” returned 33,800 records on 3 November 2026. Use LegalCheck or your own matching before you attach a case to a person.
• Pending is not guilt; disposed is a status, not an outcome. A disposed suit may have been settled, withdrawn or decreed. See the case status dictionary.
• Courts publish late. Our refresh test shows that a stale-looking date can be the court’s own gap.
• Some fields are sparse. Parsed statute citations, hearing counts on High Court records and case types on older records can be missing or marked UNKNOWN.
• AI summaries simplify. Check them against the extracted text in the same response.
• Nothing is a certified copy. For filing in court, get a certified copy from the court.

Where should you read next?

This page is the reference; these posts are the tutorials. Start with how to use the eCourtsIndia API for a guided tour and the developer quickstart for a first integration. The hub post on India public records APIs shows how court data combines with company and land records, and our comparison of court data APIs in India helps you test us against others. For company litigation checks, read finding all cases against a company. For the Acts and Sections endpoints in depth, see our Indian statute API guide.

Sources and method

• eCourtsIndia API docs, ecourtsindia.com/api/docs (endpoint groups, Case Search v2, cause list, refresh, LegalCheck, electoral, errors and timings), read 3 November 2026.
• eCourtsIndia pricing, ecourtsindia.com/api/pricing, and API page, ecourtsindia.com/api, read 3 November 2026.
• Live calls through the eCourts MCP on 3 November 2026: search_cases (no filters; HDFC Bank in Lucknow; NCLTAL0; State Bank of India in tribunals; “Rahul Sharma”), get_case_brief (DLHC010813232015), get_order_ai_analysis (order-16.pdf), refresh_and_wait_for_case, check_cnr_causelist (3 CNRs), lookup_provision (NI Act s.138) and lookup_case_provisions.
• Our own unauthenticated test of the partner API returning 401 INVALID_TOKEN, 3 November 2026.
• eCourts Services portal, services.ecourts.gov.in, and Surepass eCourts API page, surepass.io/ecourts-api, read 3 November 2026.

Frequently asked questions

Get a key, run one search on a company you know, and read what comes back.