Pakistan Case Law

Connect your AI assistant to 232,000+ Pakistani judgments

Setup guides verified 22 August 2026 · Server version 1.0.0 · Listed in the official MCP registry as com.pakistancaselaw/caselaw

This site publishes 232,000+ reported Pakistani judgments (Supreme Court, Federal Shariat Court, High Courts and tribunals, 1970 onward) as a Model Context Protocol (MCP) server. That lets an AI assistant search the corpus and walk the citation graph while you are talking to it, instead of answering from memory and inventing citations.

Free · read-only · no API key · no account. The database is opened in read-only mode, so the server is physically incapable of altering any record. Rate limit: 240 tool calls per minute per session — generous enough for a deep research run.

What you are connecting to

Endpoint URL	`https://pakistancaselaw.com/mcp`
Transport	Streamable HTTP (works with clients that ask for SSE too)
Authentication	None required
Cost	Free

Your AI calls a tool; the server reads the judgment database and returns cases with citations and links. Nothing travels the other way — the server never receives your conversation.

Which assistants can connect, and on which plan

MCP support varies a great deal between products, and the free tiers are where they differ most. Each row below was checked against the vendor's own documentation on 22 August 2026. Vendors change these terms often — if you find a discrepancy, their documentation wins, and please tell us so this page can be corrected.

Assistant	Works on a free plan?	Where you add it
Claude (web & desktop) Recommended	Yes — the Free plan allows one custom connector	Settings → Connectors
Claude Code (terminal)	Yes	A single command
Grok (grok.com)	Yes — connectors are available to all Grok users	grok.com/connectors
ChatGPT	No — needs Plus, Pro, Business, Enterprise or Edu; web only	Settings → Developer mode
Gemini Spark Recommended	No — needs Google AI Pro or AI Ultra	Gemini web → Connected Apps
Gemini CLI (terminal)	Yes	A single command
Gemini — ordinary chat (not Spark)	No — custom MCP apps run in Spark only	—
Cursor and other MCP clients	Yes	JSON config file

Our recommendation. From our own use, Claude and Gemini Spark handle legal research with these tools best — Spark for long autonomous research runs, Claude for close reading and citation-checking. Claude is also the best free option, since Spark needs a paid Google AI plan. If you want this at no cost: Claude, Grok or Gemini CLI.

Claude — web and desktop Recommended

Plans: Free, Pro, Max, Team and Enterprise. Free accounts may add one custom connector, so if you already have one you will need to remove it first.

Open Claude and go to Settings → Connectors (in some versions: Customize → Connectors).
Click the + button next to Connectors, then choose Add custom connector.
Name it Pakistan Case Law.
Paste the URL: https://pakistancaselaw.com/mcp
Leave Advanced settings (OAuth Client ID and Secret) empty — this server needs no login.
Click Add. The six caselaw_* tools should appear under the connector.

On Team and Enterprise plans an Owner must add it first, under Organization settings → Connectors → Add → Custom → Web. Members then click Connect on it.

Official documentation: Get started with custom connectors using remote MCP

Claude Code — terminal

Plans: any Claude Code account.

claude mcp add --transport http caselaw https://pakistancaselaw.com/mcp

Then confirm it registered:

claude mcp list

Inside a session, /mcp shows connection status and the tools it exposes.

Grok — grok.com

Plans: connectors are available to all Grok users. On Business and Enterprise plans a team admin must provision the connector in the cloud console before members can use it.

Go to grok.com/connectors — or click the + button in the composer and choose Connectors.
Click New Connector (or + Add connector).
Choose Custom.
Paste https://pakistancaselaw.com/mcp as the MCP server URL.
Leave authentication blank — none is required.
Save. Grok discovers the tools and offers them in conversation like any built-in connector.

Grok supports Streamable HTTP and SSE, both of which this server speaks. For programmatic use, xAI also accepts an MCP server URL in the tools array of the API.

Official documentation: Grok connectors · Remote MCP tools (API)

ChatGPT — paid plans only

Not available on the free plan. OpenAI's developer documentation lists developer mode for Plus, Pro, Business, Enterprise and Edu accounts, on the web only. Their help centre article is more restrictive still and mentions Business and Enterprise/Edu — so if you are on Plus and do not see the toggle, that is why. Both links are below; check them before subscribing for this.

In ChatGPT on the web, open Settings → Security and login (workspace admins on Business or Enterprise must enable it first under Workspace Settings → Permissions & Roles → Connected Data).
Turn on the Developer mode toggle.
Go to the Plugins / Apps area and click + to create a developer-mode app.
Enter https://pakistancaselaw.com/mcp as the MCP server URL.
Choose No Authentication — this is an accepted option and is what this server uses.
Save. The app appears under Drafts, and becomes selectable via the Developer mode tool in the composer.

Official documentation: ChatGPT developer mode (developer docs) · Developer mode and MCP apps (help centre)

Gemini Spark Recommended

Gemini Spark is Google's agentic assistant, announced at I/O in May 2026. It runs long tasks on its own in the background rather than only answering in a chat window, which suits case-law research that needs many searches and a walk through the citation graph. Spark supports custom MCP servers.

Spark is not free, and it is not ordinary Gemini chat. It launched on Google AI Ultra and expanded to Google AI Pro ($20/month) from late July 2026, starting in the US with other countries following. Custom MCP apps run inside Spark only — adding one does not give the normal Gemini chatbot access to it.

Open gemini.google.com on a computer — the connection must be made on the web, not in the mobile app.
Click Settings & help at the bottom.
Select Connected Apps.
Find Custom apps for Spark and click Add a custom app.
Enter the MCP server URL: https://pakistancaselaw.com/mcp
Click Next and follow the prompts. No credentials are needed — leave Advanced features alone, as that is only for servers requiring a login.

Once connected on the web, the app works in Spark on both the mobile and web Gemini apps.

Official documentation: Connect & manage custom apps for Gemini Spark

Gemini CLI — the free Google option

If you do not have a paid Google AI plan, the Gemini CLI connects to MCP servers free, from your terminal.

gemini mcp add --transport http caselaw https://pakistancaselaw.com/mcp

Or edit ~/.gemini/settings.json by hand — note the field is httpUrl, not url:

{
  "mcpServers": {
    "caselaw": {
      "httpUrl": "https://pakistancaselaw.com/mcp",
      "timeout": 30000
    }
  }
}

Verify with gemini mcp list — it should report Connected.

Gemini CLI reads tools only. It ignores MCP prompts and resources, so the built-in research_proposition workflow will not appear there. All six tools work normally.

Official documentation: MCP servers with Gemini CLI

Cursor and other MCP clients

Most clients accept a JSON config of this shape:

{
  "mcpServers": {
    "caselaw": {
      "type": "http",
      "url": "https://pakistancaselaw.com/mcp"
    }
  }
}

If a client offers no HTTP option and expects a local command, bridge it with the standard proxy:

npx -y mcp-remote https://pakistancaselaw.com/mcp

The six tools

All are read-only and annotated as such, so clients supporting automatic permissions can run them without prompting for every call.

Tool	What it does
`caselaw_search`	Full-text search across all 232,000+ judgments. Sort by `relevance` (nearest on the point, then court seniority, bench size, recency, citations), `newest`, `oldest` or `court`.
`caselaw_lookup_citation`	Find the judgment at an exact law-report citation — journal, year, page (for example PLD 1995 Supreme Court 34).
`caselaw_get_case`	Fetch one judgment. `section='summary'` returns metadata plus the AI headnote; `section='full'` adds the judgment text.
`caselaw_get_citations`	Walk the citation graph forward — the precedents this judgment relied on.
`caselaw_get_cited_by`	Walk it backward — later judgments citing this one, showing how it was subsequently treated.
`caselaw_most_cited`	The most-cited landmark judgments in the corpus.

The server also ships a reusable prompt, research_proposition, which lays out a multi-step research plan: re-phrase and search several ways, triage on the headnotes, then walk the citation graph in both directions. Clients that support MCP prompts will offer it; Gemini CLI will not.

Example prompts to try

“Using the Pakistan Case Law tools, find Supreme Court authority on whether bail once granted can be cancelled without strong grounds, and tell me which cases are still followed.”
“Look up PLD 1995 Supreme Court 34, summarise what it decided, then show me the later cases that cite it.”
“What are the leading Pakistani authorities on maintainability of a constitutional petition where an alternate remedy exists? Give me the contrary authority as well.”
“Trace the line of authority on khula and the wife's right to dissolution — start from the most-cited case and work forward to the current position.”

If it does not connect

Symptom	Likely cause
Claude says you have reached your connector limit	You are on the Free plan, which allows one custom connector. Remove the other one.
No Developer mode toggle in ChatGPT	Your plan does not include it, or a workspace admin has not enabled it.
Tools listed but every call errors	You may be rate limited at 240 calls per minute per session. The reply says so and sets `Retry-After: 60` — wait a minute.
Client rejects the URL	Check it is exactly `https://pakistancaselaw.com/mcp` — https, no trailing slash.
No prompts or resources appear	Expected on some clients (Gemini CLI reads tools only). The tools still work.

Still stuck? Email contact@pakistancaselaw.com with the client you are using and what you see.

How to rely on the results

Every result carries its citation and a link to the judgment page on this site, so any answer can be checked against the source.
Each judgment records the court's own disposition — appeal dismissed, bail granted, and so on. A judgment can state the principle you rely on while the applicant still lost on the facts, so always read the outcome alongside the holding.
AI headnotes cover part of the corpus and are growing. Their absence says nothing about a case's importance.
The headnotes are machine-generated summaries — a triage aid, not a substitute for reading the judgment.

This is research assistance, not legal advice. AI assistants can misread a judgment or overstate what it decided. Verify every authority against the official law report — PLD, SCMR, CLC, YLR, PLC, MLD and others — before relying on it in practice.

Data and privacy

The server requires no account and collects no personal data. It receives only the search terms or judgment identifiers your AI client sends in a tool call. It does not receive your conversation, your files, or your chat history, and it cannot request them. It records anonymous usage counts and a short-lived per-IP request count to enforce the rate limit. The judgments themselves are public documents of the courts. Full Privacy Policy.

About

Pakistan Case Law is built and maintained by Hamadullah Shah, an advocate practising in Pakistan. It began as a working lawyer's answer to a practical problem: Pakistani case law is scattered across law reports that are hard to search and often paywalled. The corpus, the citation graph, the AI headnotes and this MCP server are all free to use.

Contact and professional profile: LinkedIn · contact@pakistancaselaw.com

Disclaimer Privacy Policy Terms of Use llms.txt Search