Inspiration
Legal documents are often written for lawyers, not ordinary people. We wanted to make it easier for people to understand what a document actually means, which clauses deserve attention, and what they can do next without having to decode pages of legal language.
What it does
CivicShield analyzes legal documents and identifies potentially important or concerning clauses. It connects each finding to relevant legal sources, explains why it matters, shows the original clause, highlights the evidence behind the analysis, and provides practical next steps and response generation.
Our core flow is:
Document → Evidence → Law → Risk → Action
How we built it
CivicShield is built as a modern web application with a Next.js, React, TypeScript, and Tailwind CSS frontend. We built a document-analysis pipeline for extracting and classifying clauses, detecting potential legal issues, retrieving relevant legal sources, and generating evidence-backed explanations and action plans.
We also created a deterministic Demo Mode so the complete experience can be demonstrated reliably without depending entirely on external AI services.
Challenges we ran into
The biggest challenge was making legal AI useful without turning it into a generic chatbot. We needed to make every finding traceable to the original document and relevant sources while clearly separating document facts, AI reasoning, and legal information.
We also had to handle document extraction, different file formats, uncertain conclusions, and reliable demo behavior.
Accomplishments that we're proud of
Built an end-to-end legal document analysis workflow. Created clause-level findings with document highlighting. Added an evidence trail connecting findings to legal sources. Built practical action plans instead of only providing explanations. Added response generation based on identified issues. Created a polished, judge-ready TenantShield experience. Designed Demo Mode for a reliable live demonstration.
What we learned
We learned that building legal AI isn't just about generating a good answer. Trust, traceability, and context matter just as much as intelligence.
A useful system needs to show users why it reached a conclusion, where the information came from, what remains uncertain, and what they can actually do with the result.
What's next for CivicShield
We plan to expand CivicShield beyond rental agreements into other high-impact legal and civic workflows, while improving its legal-source coverage, document understanding, and evidence tracing.
Our long-term goal is to make CivicShield a trusted bridge between complex legal information and practical action—helping people understand not just what the law says, but what it means for their specific situation.
Built With
- ai/rag
- legalsourceretrieval
- lucide-react
- mammoth
- next.js
- node.js
- pdf-parse
- react
- tailwind-css
- typescript
- zod
