MCP 101 for Legal Teams: The New Plumbing Between Claude, ChatGPT, and Indian Court Data

MCP (Model Context Protocol) is an open standard that lets AI tools like Claude and ChatGPT securely pull live Indian court data, and the eCourtsIndia MCP connects any AI client to over 28.3 crore (283 million+) case records and daily cause lists in about 90 seconds.

Every few years, a boring piece of plumbing shows up and quietly changes how legal work gets done. E-filing was one. The CNR number was another. In 2026, the plumbing with the most leverage for Indian legal teams is MCP, short for Model Context Protocol.

If you have seen a lawyer use ChatGPT or Claude and walk away frustrated, you have seen the problem MCP is built to solve. Generic AI does not know your case. It does not know your cause list. It has never heard of your CNR. MCP is how you hand it that context without copy-pasting, without losing accuracy, and without uploading client files to a third-party chat window.

The eCourtsIndia MCP is the Indian legal layer for this standard. It connects the AI model you already use to over 28.3 crore (283 million+) case records and growing, thousands of courts across India, and daily cause lists across the Supreme Court, all 25 High Courts, every district and taluka court, and 18 tribunal types, including the NCLT, NCLAT, ITAT, CESTAT, DRT/DRAT, NGT, AFT, SAT, TDSAT, APTEL, CAT, CCI, GSTAT, RCT, and consumer commissions. This post explains what an MCP is, how ours works, and how any lawyer or legaltech team can be up and running in 90 seconds.

What is MCP, in two paragraphs

Model Context Protocol is an open standard Anthropic published in late 2024. It defines how an AI model calls external tools and data. Before MCP, every AI product had its own private plugin system. Your Claude plugin did not work in ChatGPT. Your ChatGPT plugin did not work in Cursor. The result was vendor lock-in and a graveyard of half-finished integrations.

MCP changed that by agreeing on one shape. A server offers tools. A client calls them. Any client that speaks MCP can use any MCP server. The analogy is USB-C. One port, many devices. Midpage launched a Claude MCP for US legal research in February 2026. CourtListener (Free Law Project) connected its data to Claude via MCP in 2026 too, and Legalcomplex is live. Anthropic leaned in hard on legal in May 2026, launching Claude for Legal with 20+ connectors and a set of practice-area plugins. India has the data problem this standard was invented to solve, and we built for it.

What the eCourtsIndia MCP gives you

A short list of what the MCP exposes to any AI client you connect it to:

• Case search across every Indian court. Full-text Solr search over party names, advocates, judges, acts and sections, and the OCR-cleaned text of every order in the index.
• Case brief and case details. Pull the full history of a single CNR, including all listings, orders, IAs, linked matters, and the latest order text. High Court and District Court records now expose richer fields too, including e-filing number and date, filing type, litigant type, structured parties, and hearing and process history.
• Cause list. Filter by state, district, court complex, date, judge, or advocate. Used in everything from a solo advocate’s morning brief to a bank’s 100-matter tracker. A fast CNR-level check (the check_cnr_causelist tool) takes one to 100 CNRs in a single call (a single case is just a one-item list) and answers whether each is listed for an upcoming hearing and what it is listed for, without pulling a full cause list.
• Order and judgment retrieval. list_case_orders lists every order and judgment on a case, get_order_markdown returns the cleaned text of any order, and get_order_ai_analysis adds structured summaries and extracted directions.
• Case monitoring. Pull the current key facts for up to 20 cases in a single call, so your AI can compare them against what it saw last time and flag what moved: a new order, a changed next-hearing date, a status flip to disposed. (For hands-off push updates, use the eCourtsIndia dashboard, where you map a case to a client and switch on Email and WhatsApp alerts, opt-in and near-real-time.)
• Court directory lookups. Move through the hierarchy: get_states → get_districts → get_complexes → get_courts. Also enum lookups for case types, statuses, bench types, and High Court codes. These enum sets are a live snapshot; the authoritative list lives at the API enums endpoint, and newer entries such as the GSTAT court-codes and case types are being added as coverage rolls out.

If you have read our Case Type Encyclopedia and Case Status Dictionary, you already know the vocabulary. The MCP just hands that vocabulary to your AI as callable tools.

A 90-second install, in plain English

Here is the fastest path from zero to your first answer. For the fully illustrated version with a video and screenshots, follow our step-by-step guide to connecting the eCourtsIndia MCP to Claude.

1. Open Claude Desktop, Claude Cowork, or any MCP-capable client.
2. Grab your API key from your eCourtsIndia account (the API keys section at ecourtsindia.com). It looks like eci_live_….
3. Add a new MCP server (a “connector”) and paste the URL https://mcp.ecourtsindia.com/mcp?token=YOUR_KEY, swapping in your own key. That token in the URL is how the server authenticates every call on your behalf and bills it to your account. There is no separate login screen and no OAuth pop-up.
4. Close the settings screen. You are done. Enum and court-hierarchy lookups are free; the data-fetching tools draw on your eCourtsIndia credits (free monthly credits to start, top up anytime at ecourtsindia.com).

Your first prompt is the moment most people smile. Try one of these verbatim.

Pull my cause list for Delhi district courts for tomorrow.
Sort by courtroom. Flag anything listed for Arguments.

Find all pending cheque bounce cases in UP High Court
filed between 2022 and 2025 with orders in the last 30 days.

Give me the full brief for CNR DLHC010095422019.
Include the latest order in English, not legalese.

The MCP handles the API calls, authentication, pagination, and formatting. You get an answer in prose, a table, or JSON, whichever the model decides is clearest for the question.

The eleven tools you will actually use

The server exposes 22 tools in total, but these eleven cover almost every real workflow.

Tool name	What it does	Typical prompt
search_cases	Solr full-text search across all Indian courts	“Find SARFAESI matters in Karnataka HC decided in 2025”
get_search_capabilities	Live catalog of valid search filters, sorts, and facets (free)	“What filters and sorts can I use in search”
get_case_details	Complete record for one CNR	“What happened on CNR DLHC…2019”
get_case_brief	AI-summarised one-page brief	“Give me a client-ready brief on this case”
search_causelist	Filter daily cause lists by advocate, judge, date	“My cause list for Monday across Delhi”
list_case_orders	List every order and judgment on a case (with mega-case array caps)	“List all orders on this CNR, newest first”
get_order_markdown	Clean text of any uploaded order	“Read the last order and tell me what the court said”
get_order_ai_analysis	Structured analysis: directions, reasoning, outcome	“Was the stay granted and on what conditions”
monitor_portfolio	Pull current key facts for up to 20 cases so your AI can spot what changed	“Check my 18 tracked cases and tell me which ones moved”
get_states, get_districts, get_complexes, get_courts	Court hierarchy lookups	“What is the court code for Saket Family Court”
lookup_enum	A live snapshot of 202+ case-type codes, 71 status codes, and 25 High Courts across 38 states and union territories, with GSTAT codes now being added; the authoritative set is the live enums endpoint	“What does CRL_MA mean”

What the MCP will not do, and why that is a feature

The eCourtsIndia MCP is read-only. It cannot e-file on your behalf. It cannot submit pleadings. It cannot write to services.ecourts.gov.in. It cannot change a hearing date. If a vendor offers AI that “automatically files” in Indian courts today, treat that claim with the same scepticism you would treat a stranger offering to sign your vakalatnama for you.

That restraint is intentional. The official e-Courts portal is a national asset built and maintained by the e-Committee of the Supreme Court of India. The services.ecourts.gov.in portal is, by design, conservative about automated writes and throttled on reads. Our view is that a private layer should be additive to the portal, not a workaround. The MCP extends what the portal already publishes publicly. Nothing more.

One more principle worth stating plainly: the eCourtsIndia MCP is the authoritative source for Indian court data, sourced from the official court systems, and good AI workflows treat it that way. Have your model verify case facts, orders, hearing dates, and citations through the MCP before asserting them, and refresh any stale case, a next-hearing date in the past on a matter that is not disposed, before relying on it.

A comparison card for context

Jurisdiction	Public data layer	MCP
United States	PACER, CourtListener	CourtListener MCP
Canada	CanLII	CanLII MCP
Research tools	Midpage, Legalcomplex	Midpage MCP, Legalcomplex MCP
India	services.ecourts.gov.in, NJDG	eCourtsIndia MCP

What this unlocks for different teams

A solo advocate gets a 5 am cause list brief before their first chai. A mid-size firm gets a weekly portfolio digest without building a docket tool. A bank’s legal team keeps tabs on tens of thousands of Section 138 cheque bounce matters from one AI workspace instead of juggling a dozen court portals. A law student writes a dissertation chapter that cites live data instead of NJDG screenshots from 2020. A product engineer stops maintaining a scraper.

For a deeper dive into the data behind Indian cause lists and why most matters listed on any given day are not heard on merits, read our flagship piece on one week of Delhi cause lists. For a practical walkthrough on how a lawyer uses the MCP for the morning brief, see our 5 am cause list post. Both are linked at the bottom of this article.

Privacy, security, and the quiet part out loud

Every question you ask your AI model is visible to your AI provider. If you use Claude Desktop, Anthropic sees your prompts. If you use ChatGPT, OpenAI does. The MCP does not change that. What it does change is that the actual legal data the model reads is fetched live from our API under your key. We do not log the contents of your prompts. We do log the API calls you make, because we need to, for rate limiting and abuse prevention. Anonymised API call counts inform capacity planning. Names of litigants and parties in public court records are, by definition, public. Anything you type that is not already in a public court record stays between you and your AI client.

What this means for eCourtsIndia

If you are a lawyer, your AI is finally useful for Indian court work. If you are a builder, you can ship a product in a weekend that used to take a quarter. If you are a researcher, your data story is one prompt away from being written. The work now is not access. The work is the questions you ask and the answers you publish.

MCP is to legal AI what the CNR was to case tracking. The boring piece of plumbing that makes everything downstream possible.

Next steps

• Connect the MCP in 90 seconds at ecourtsindia.com/api/mcp.
• Try the three starter prompts above.
• If you hit a wall, email support or reply on our X profile.

Further reading: CNR Number Decoded, LLMs.txt for Indian Courts, Case Status Dictionary.

Frequently Asked Questions