Answers you can check
ChallengeA legal answer without sources is worse than no answer.
How we solved itCase-law overviews are written only from retrieved judgments, with inline citations — anything uncited is dropped.
An AI legal assistant for Pakistani law — research with cited judgments, documents drafted with AI, contracts checked for risk, and a marketplace of verified lawyers. In English and Urdu.
See it livepakistanlawbot.com Legal drafting Document analysis Lawyer marketplace
“No one should face a legal challenge alone because they can’t afford a lawyer or don’t know where to turn.” LawBot’s mission is to make justice accessible for every Pakistani — legal research, drafting and a verified lawyer, in plain English or Urdu.
The public, lawyers, law firms and law students each get tools built for how they work — on one API and one legal database.
People describe their problem in their own words; LawBot suggests the right kind of lawyer and books a consultation with a verified advocate.
6 core featuresTools for the people who practise and study law every day.
5 core featuresFrom a question in English or Urdu to an answer you can check.
Type or speak — in English, Urdu or Roman Urdu.
Keyword and vector search across judgments, filtered by court and outcome.
The model writes only from the judgments it retrieved, citing each one.
Judgments the answer doesn’t cite are dropped from the result.
Translate to Urdu, draft a document or book a verified lawyer.
A pre-rendered React app talks to a Node.js API; AI work runs in a separate Python service, and MongoDB Atlas holds both the app data and the legal search indexes.
Web appReact · Vite · pre-rendered for SEO
APINode.js · Express · TypeScript — auth, quotas, drafting, case law and the marketplace
Legal data & searchMongoDB Atlas Vector Search + Atlas Search
AI servicesPython AI server · Claude · GPT · Gemini
PaymentsStripe (PKR) · Apple & Google in-app
MessagingWhatsApp Cloud API · emailChallengeA legal answer without sources is worse than no answer.
How we solved itCase-law overviews are written only from retrieved judgments, with inline citations — anything uncited is dropped.
ChallengePeople ask in English, Urdu and Roman Urdu.
How we solved itMultilingual embeddings put English and Urdu in one vector space, running locally at no per-query cost.
ChallengeAn open text box will be asked for anything.
How we solved itA fast guard model refuses non-legal requests before Claude drafts, and every draft carries an advocate-review notice.
ChallengePeople need to know a lawyer is real before they book.
How we solved itBar-licence checks by the LawBot team, a WhatsApp code before booking, and reminders for every appointment.
The product film, and the document analyser at work.
Proven, well-supported tools — chosen so the platform is easy to run and to extend.
ReactWeb app
TypeScriptFront end & API
ViteBuild & pre-rendering
Tailwind CSSStylingNode.jsAPI
ExpressAPI frameworkMongoDBData, vector & keyword searchPythonAI services
AnthropicLegal drafting
OpenAICase-law overviews & matching
Google GeminiJudgment headnotes
Hugging FaceMultilingual embeddingsStripePaymentsWhatsAppCodes & reminders
CloudflareEdge & security
SentryError trackingLaw demands sources, human review and care with private documents. LawBot is built around all three.
Answers point to the judgment or statute they came from.
Drafts are marked for review by an enrolled advocate.
Requests outside the legal domain are refused before drafting.
English, Urdu and Roman Urdu, by text or voice.
An MCP server lets outside assistants use LawBot’s tools.
Live today: research, drafting, document analysis and the lawyer marketplace all run at pakistanlawbot.com. Answers are legal information, not legal advice. Visit LawBot
Tell us who will use it and what it must do. We’ll map the first release, the stack and the right team — in a free 30-minute strategy call.
