Inspiration
Filing patents is one of the most important ways to protect innovation, yet the process is slow, costly, and often limited to large corporations with legal teams. Startups, independent inventors, and researchers struggle to navigate complex legal requirements, which delays or prevents them from protecting their ideas. We asked: what if AI agents could take on the repetitive, expensive, and time-consuming work of patent drafting and compliance — so anyone can protect their innovation?
What it does
Agentic Law Firm: AI Patent Agent is a multi-agent system that automates patent drafting and submission workflows.
- Analyst Agent: drafts claims and descriptions
- Skeptic Agent: challenges validity and checks prior art
- Judge Agent: finalizes the draft and ensures compliance
- Integrates patent databases (USPTO/EPO) and web search to cross-reference existing filings
- Produces submission-ready documents that reduce cost and time
In short: Agentic Law Firm transforms hours of expensive legal work into minutes of automated, AI-powered drafting.
How we built it
- Backend: FastAPI to coordinate multi-agent flows
- Models: GPT-5 for reasoning, plus fine-tuned legal LLMs for compliance
- Search: USPTO/EPO APIs + Brave Search for prior art retrieval
- Memory: Neo4j graph for storing prior cases and references
- Frontend: Next.js app for uploading invention details and exporting draft filings
Challenges we ran into
- Designing agents that reason like lawyers without hallucinating
- Ensuring outputs met strict patent office compliance requirements
- Balancing real-time usability with the depth of legal checks
- Creating explainability features so lawyers trust the AI’s verdicts
Accomplishments that we're proud of
- Built a working patent draft generator in under a week
- Automated prior art search using real patent databases
- Reduced drafting time by over 80% in prototype tests
- Designed a developer- and lawyer-friendly API for rapid adoption
What we learned
- Multi-agent orchestration is powerful for self-checking legal workflows
- Legal automation must combine accuracy, transparency, and compliance
- Inventors and startups are eager for tools that lower the barrier to IP protection
- Trust is as important as speed — outputs must be explainable to gain adoption
What's next for Agentic Law Firm: AI Patent Agent
- Fine-tune on patent corpora to improve legal accuracy
- Expand into contracts, compliance, and litigation support
- Partner with law firms and accelerators to pilot real-world use cases
- Open-source SDK to let startups integrate patent drafting into their workflows
- Ultimately: become the AI legal infrastructure layer, making law faster, fairer, and accessible to all
Built With
- amazon-web-services
- brave-search-api
- docker
- fastapi
- github
- gpt-5
- javascript
- legal-fine-tuned-llms
- neo4j
- next.js
- postgresql
- python
- react
- typescript
- uspto/epo-apis
Log in or sign up for Devpost to join the conversation.